CN117010413A - Community question and answer method, device, storage medium and computer equipment - Google Patents

Abstract

This application discloses a community question and answer method, which includes: displaying a question and answer interface, which includes a target question and a first control; the target question is a posted question or a follow-up question for a target answer, and the target answer is an answer for the posted question; in response to Trigger the first control to obtain the reply information of the target question. The reply information is semantically encoded and decoded based on the target question; the reply information is displayed in association with the target question, and the first mark is displayed for the reply information. The first Flag is used to indicate that the reply message is automatically generated. This application applies artificial intelligence technology to achieve automatic replies to target questions in the community by semantically encoding and decoding the target questions to obtain the corresponding reply information, avoiding asking users to passively wait for long periods of time for other users to respond to target questions. This improves the timeliness of replying to questions in the Q&A community.

Description

Community question and answer method and device, storage medium and computer equipment

Technical Field

The application relates to the technical field of artificial intelligence, in particular to a community question-answering method, a device, a storage medium and computer equipment.

Background

The question-answering community is an internet platform for providing free communication among a plurality of users, and is a network site for communication, learning and information sharing by applying new internet technology. In the question-answering community, people can put out questions on line, discuss and communicate with each other, so that quick service can be provided for users, and the service quality of the Internet is improved.

The question and answer community platform supports numerous interactions in the form of user communities, such as sharing articles, asking questions, answering questions and comments, etc., for example, discussing specific technical topics or questions with peers and experts, which allow users to efficiently and correctly communicate and communicate with each other about the technology or questions they are interested in.

Currently, users can ask questions about topics that the users want to discuss in a question and answer community, and then wait for other users to answer to the questions. However, the period of passively waiting for other users to answer questions is long, so that the timeliness of the answer of the questions is low.

Disclosure of Invention

The embodiment of the application provides a community question and answer method, a device, a storage medium and computer equipment, which are used for solving the problems that in the related technology, the period of passively waiting for other users to answer questions in a question and answer community is long, so that the timeliness of question reply is low.

In one aspect, an embodiment of the present application provides a community question-answering method, which includes: displaying a question-answer interface, wherein the question-answer interface comprises a target question and a first control; the target question is a release question or an inquiry question aiming at a target answer, and the target answer is an answer aiming at the release question; responding to the triggering operation of the first control, and acquiring reply information of the target problem, wherein the reply information is obtained by carrying out semantic coding and decoding on the basis of the target problem; and displaying the reply information in association with the target problem, and displaying a first mark aiming at the reply information, wherein the first mark is used for indicating that the reply information is automatically generated.

According to the embodiment of the application, the first control in the question-answer community can be responded to the triggering operation of the first control in the question-answer interface, the questions can be issued or the questions can be answered automatically by the user, the first mark is displayed for the automatically generated reply information, so that the user can distinguish the reply information obtained by the working answer of the person and the automatically generated reply information based on the first mark, the first control for the target questions can be selectively triggered by the user, the target questions can be subjected to semantic coding and then be decoded to automatically generate the reply information of the target questions, the problem can be prevented from being passively waited for a long time to reply the questions manually, the period of the reply of the questions is effectively shortened, and the timeliness of the reply is improved. In addition, in the scheme of the application, the reply information obtained by carrying out semantic encoding and decoding based on the target problem can be ensured to be obtained by decoding under the condition of fully understanding the semantics of the target problem, so that the accuracy of the reply information obtained aiming at the target problem can be ensured.

On the other hand, the embodiment of the application provides a community question-answering method which is applied to a server and comprises the following steps: receiving a reply request sent by a client, wherein the reply request is generated by the client in response to the triggering operation of a first control in a question-answer interface, the question-answer interface further comprises a target question, the target question is a release question or a query question aiming at a target answer, and the target answer is an answer aiming at the release question; responding to the reply request, carrying out semantic coding and decoding based on the target problem, and obtaining reply information of the target problem; and sending the reply information of the target problem to the client so that the client can correlate and display the reply information with the target problem, and displaying a first mark aiming at the reply information, wherein the first mark is used for indicating that the reply information is automatically generated.

On the other hand, the embodiment of the application also provides a community question-answering device, which comprises: the question and answer display module is used for displaying a question and answer interface, and the question and answer interface comprises a target question and a first control; the target question is a release question or an inquiry question aiming at a target answer, and the target answer is an answer aiming at the release question; the information acquisition module is used for responding to the triggering operation of the first control to acquire the reply information of the target problem, wherein the reply information is obtained by carrying out semantic coding and decoding on the basis of the target problem; and the reply display module is used for carrying out associated display on the reply information and the target problem, and displaying a first mark aiming at the reply information, wherein the first mark is used for indicating that the reply information is automatically generated.

On the other hand, the embodiment of the application also provides a community question-answering device which is applied to the server and comprises: the request receiving module is used for receiving a reply request sent by the client, wherein the reply request is generated by the client in response to the triggering operation of the first control in the question-answer interface, the question-answer interface further comprises a target question, the target question is a release question or an inquiry question aiming at a target answer, and the target answer is an answer aiming at the release question; the reply generation module is used for responding to the reply request, carrying out semantic coding and decoding based on the target problem and obtaining reply information of the target problem; the reply sending module is used for sending reply information of the target problem to the client so that the client can display the reply information in association with the target problem, and a first mark is displayed for the reply information, wherein the first mark is used for indicating that the reply information is automatically generated.

In another aspect, an embodiment of the present application further provides a computer readable storage medium storing a computer program, where the community question-answering method is executed when the computer program is executed by a processor.

On the other hand, the embodiment of the application also provides a computer device, which comprises a processor and a memory, wherein the memory stores a computer program, and the computer program executes the community question-answering method when being called by the processor.

In another aspect, embodiments of the present application also provide a computer program product comprising a computer program stored in a storage medium; the processor of the computer device reads the computer program from the storage medium, and the processor executes the computer program to cause the computer device to execute the steps in the community question-answering method described above.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are needed in the description of the embodiments will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present application, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.

Fig. 1 shows a schematic diagram of a system framework according to an embodiment of the present application.

Fig. 2 shows another system frame schematic provided by an embodiment of the present application.

Fig. 3 shows a flow chart of a community question-answering method according to an embodiment of the present application.

Fig. 4 shows an interface schematic diagram of a creation interface provided by an embodiment of the present application.

Fig. 5 shows an interface schematic diagram of a question-answer interface according to an embodiment of the present application.

Fig. 6 shows an interface schematic diagram of another question-answer interface provided by an embodiment of the present application.

Fig. 7 shows an interface schematic diagram of another question-answering interface provided by an embodiment of the present application.

FIG. 8 is an interface diagram of yet another problem interface provided by an embodiment of the present application.

FIG. 9 shows an exemplary diagram of a control response provided by an embodiment of the present application.

FIG. 10 illustrates another control response example diagram provided by an embodiment of the present application.

Fig. 11 is a schematic diagram of an interface for displaying reply information according to an embodiment of the present application.

Fig. 12 is a schematic diagram of another interface for displaying reply information according to an embodiment of the present application.

Fig. 13 is a schematic diagram of another interface for displaying reply information according to an embodiment of the present application.

Fig. 14 is a schematic flow chart of another community question-answering method according to an embodiment of the present application.

Fig. 15 shows a problem issuing flowchart provided by an embodiment of the present application.

Fig. 16 shows a flowchart of a validity check provided by an embodiment of the present application.

Fig. 17 shows a flowchart of text preprocessing according to an embodiment of the present application.

FIG. 18 shows a flow chart of a semantic encoding provided by an embodiment of the present application.

Fig. 19 shows a flowchart of yet another community question-answering method according to an embodiment of the present application.

Fig. 20 shows a flowchart of a decoding process provided by an embodiment of the present application.

Fig. 21 shows a flowchart of still another community question-answering method according to an embodiment of the present application.

Fig. 22 shows a timing chart of a community question-answering method according to an embodiment of the present application.

Fig. 23 shows a flowchart of reply message generation according to an embodiment of the present application.

Fig. 24 is a block diagram of a community question-answering device according to an embodiment of the present application.

Fig. 25 is a block diagram of another community question-answering device according to an embodiment of the present application.

Fig. 26 is a block diagram of a computer device according to an embodiment of the present application.

Fig. 27 is a block diagram of a computer-readable storage medium according to an embodiment of the present application.

Detailed Description

Embodiments of the present application are described in detail below, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to like or similar elements or elements having like or similar functions throughout. The embodiments described below by referring to the drawings are exemplary only for explaining the present application and are not to be construed as limiting the present application.

In some of the processes described in the specification, claims and drawings above, a number of steps occurring in a particular order are included, but it should be understood that the steps may be performed out of order or performed in parallel, the sequence numbers of the steps merely being used to distinguish between the various steps, the sequence numbers themselves not representing any order of execution. Furthermore, the descriptions of "first" and "second" and the like herein are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order.

In order to enable those skilled in the art to better understand the solution of the present application, the following description will make clear and complete descriptions of the technical solution of the present application in the embodiments of the present application with reference to the accompanying drawings. It will be apparent that the described embodiments are only some, but not all, embodiments of the application. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.

It should be noted that, in the specific embodiment of the present application, related data such as issue questions, inquiry questions, and reply messages are required to obtain user permission or consent when applied to specific products or technologies of the embodiments of the present application, and the collection, use, and processing of related data are required to comply with related laws and regulations and standards of related countries and regions, and the subsequent data use and processing actions are performed within the authorized range of laws and regulations and personal information subjects.

The community question-answering method of the application relates to an artificial intelligence (Artificial Intelligence, AI) technology, which is a theory, a method, a technology and an application system for simulating, extending and expanding human intelligence by using a digital computer or a machine controlled by the digital computer, sensing environment, acquiring knowledge and using the knowledge to acquire an optimal result. In other words, artificial intelligence is an integrated technology of computer science that attempts to understand the essence of intelligence and to produce a new intelligent machine that can react in a similar way to human intelligence. Artificial intelligence, i.e. research on design principles and implementation methods of various intelligent machines, enables the machines to have functions of sensing, reasoning and decision.

The artificial intelligence technology is a comprehensive subject, and relates to the technology with wide fields, namely the technology with a hardware level and the technology with a software level. Artificial intelligence infrastructure technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technologies, operation/interaction systems, mechatronics, and the like. The artificial intelligence software technology mainly comprises a computer vision technology, a voice processing technology, a natural language processing technology, machine learning/deep learning and other directions.

Among them, natural language processing (Nature Language processing, NLP) is an important direction in the fields of computer science and artificial intelligence. It is studying various theories and methods that enable effective communication between a person and a computer in natural language. The natural language processing relates to natural language, namely the language used by people in daily life, and is closely researched with linguistics; and also to computer science and mathematics. Natural language processing techniques typically include text processing, semantic understanding, machine translation, robotic questions and answers, knowledge graph techniques, and the like. For example, in the embodiment of the present application, word embedding is performed on the text of the question posed by the user, and semantic analysis is performed on the question.

Currently, users in community question and answer are mostly answering, inviting to answer or pursuing questions. For example, a user searches for a question, if there is no suitable answer to the question or no related question, a question may be created and then awaits other user answers. If the user finds the same question, then pick a useful answer to watch or ask, if not answer to the question, then pay attention to the question to wait for the answer. In addition, if the same question is not answered, a response invitation can be sent to the hot respondent who has answered the similar question, and the answer can be awaited.

Since the answer time of the respondent is unknown, the questioning user can only wait for other users to answer the question until a suitable answer appears. However, in practical application processes, users often want to get proper answers. In addition, if the respondent has a severe personal emotional color, an opinion conflict may occur with the asking user, making the answer ineffective. Therefore, the period of getting a satisfactory answer is long for the asking user, and may be even remote for an indefinite period, making the question discussion not time-efficient. In order to solve the problems, the inventor provides a community question-answering method provided by the embodiment of the application through researches.

The architecture of the system of the community question-answering method related to the application is described first.

As shown in fig. 1, the community question-answering method provided by the embodiment of the application can be applied to a system 100, and the system 100 can be used for model training. Wherein, the data acquisition device 110 is configured to acquire training data, and the training data includes a plurality of training samples. For the community question-answering method of the embodiment of the application, the training sample can be a question-answer pair formed by a question text and an answer text. For example, for a game community, after the player authorization data is acquired, a great number of questions and answers about a game policy issued by the player in the community content can be acquired, and further, preprocessing such as data cleaning is performed on the questions and answers to obtain corresponding training data, where the training data can be used for training the target model 101 for automatically replying questions and answers to the game community. After the data acquisition device 110 acquires the training data, the training data may be stored in the database 120, and the training device 130 may train to obtain the target model 101 based on the training data maintained in the database 120.

Specifically, the training device 130 may train the preset neural network based on the input training data until the preset neural network meets a preset training end condition, to obtain the trained target model 101. The preset conditions may be: the loss value of the target loss function is smaller than a preset value, the loss value of the target loss function is not changed any more, or the training times reach the preset times, and the like. The object model 101 may be used to automatically generate answers to an input question (e.g., an object question in the present application) based on the question, and the process involved in the object model 101 may include: word embedding, semantic analysis, decoding, etc. The object model 101 in the embodiment of the present application may be a deep neural network (Deep Neural Network, DNN), for example, a transducer, etc., which is not limited herein.

In an actual application scenario, the training data maintained in the database 120 is not necessarily all from the data acquisition device 110, but may be received from other devices, for example, the execution device 140 may also be used as a data acquisition end, and the acquired data is used as new training data and stored in the database 120. In addition, the training device 130 does not need to train the preset neural network based on the training data maintained by the database 120, and may train the preset neural network based on the training data obtained from the cloud or other devices, for example, when the executing device 140 is a terminal where the client is located, the obtained question-answer data of the user in the question-answer community may be used as the training data, which should not be taken as a limitation of the embodiments of the present application.

The object model 101 trained according to the training apparatus 130 described above may be applied to different systems or apparatuses, such as the execution apparatus 140 shown in fig. 1. The training device 130 and the executing device 140 may be servers or terminals, and the servers may be independent physical servers, or may be server clusters or distributed systems formed by a plurality of physical servers, or may be cloud servers for providing cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDNs, blockchains, basic cloud computing services such as big data and artificial intelligence platforms. The terminal may be a smart phone, tablet computer, notebook computer, desktop computer, etc.

In the process of performing the related processes such as calculation by the processing module 141 of the execution device 140, the execution device 140 may call the data, the program, etc. in the data storage system 150 for the corresponding calculation process, and store the data and the instructions such as the processing result obtained by the calculation process in the data storage system 150. The training device 130 may generate respective target models 101 based on different training data for different targets or different tasks, which respective target models 101 may be used to complete respective community question-and-answer tasks.

Illustratively, the training device 130 in the system 100 shown in fig. 1 may be a cloud server deployed for a service provider, and the execution device 140 may be a terminal (e.g., a smartphone) used by a user. The cloud server may perform web training on, i.e., training data, based on questions and answers to obtain a question and answer network for performing community question and answer tasks, which may include a word embedding sub-network, a semantic analysis sub-network, and various decoders. Furthermore, the terminal can deploy a training obtaining question-answering network, namely the target model to execute community question-answering tasks.

For example, when the question-answer community is a game question-answer community, when the player a issues a game policy related problem at the game community client, by clicking an "AI reply" control (Virtual Button) on a question-answer interface displayed by the game community client, the game community client may send a request to a cloud server of a game community service background, and after receiving the request, the cloud server may generate corresponding reply information, that is, a game policy answer, for the game policy related problem based on the question-answer network, and return the game policy answer to the game community client, so that the game policy answer is directly displayed at the question-answer interface by the game community client.

It should be noted that fig. 1 is only a schematic diagram of a system provided by an embodiment of the present application, and the architecture and application scenario of the system described in the embodiment of the present application are for more clearly describing the technical solution of the embodiment of the present application, and do not constitute a limitation on the technical solution provided by the embodiment of the present application. For example, the question-answering community can be a topic discussion and social communication platform such as a scientific creation community, a content sharing community, a friend-making community, a game sharing community, a learning sharing community and the like, and is not limited herein. The execution device 140 in fig. 1 may be a terminal used by a user, in other cases, the execution device 140 may also be a cloud server deployed by a service provider.

As shown in fig. 2, the community question-answering method provided by the embodiment of the application can also be applied to the system 200. Illustratively, the functions and application scenarios possessed by the data acquisition device 210, the database 220, the training device 230, and the database system 250 in the system 200 may be the same as the data acquisition device 110, the database 120, the training device 130, and the database system 150 in the system 100. The execution device 240 in the system 200 may be a cloud execution server that deploys a question-and-answer network trained by a cloud training server (i.e., training device 230) and may operate the question-and-answer network to perform community question-and-answer tasks in conjunction with the client device 260.

For example, a user may install and use a community question-answering client on a notebook computer (i.e., client device 260). When a user creates a problem at a community question-answering client, the notebook computer can send a request message to a cloud execution server through a network. Further, the cloud execution server generates a plurality of answers based on the questions created by the user using the question-answering network when receiving the request message. Further, the cloud execution server may send the filtered target answers to a community question-and-answer client of the notebook computer, and the community question-and-answer client displays the target answers so as to complete automatic answer to the user questions.

Referring to fig. 3, fig. 3 is a flow chart illustrating a community question-answering method according to an embodiment of the present application. In this embodiment, the community question-answering method may be performed by a terminal having at least functions of display, storage, calculation, and communication. As shown in fig. 3, the community question-answering method specifically includes the following steps:

step 110: displaying a question-answer interface, wherein the question-answer interface comprises a target question and a first control; the target question is a question of issue or a question of inquiry for a target answer, which is an answer for the issue.

Wherein, the question-answering interface refers to an interface for answering the target question. The issue question is a question issued by a user after the creation of a question-answer community. Optionally, after the user completes creating the issue problem at the client of the question-answering community, the issue problem may be sent to the server of the question-answering community, and then the server of the question-answering community may perform validity check on the issue problem, and issue the issue problem after the issue problem passes the validity check.

As an implementation, after a user logs in to a client of a question-and-answer community on a terminal, a question may be entered at the creation interface, which may be in text form, and further, may include a picture or video. Optionally, one or more question labels may be selected for the inputted questions, where the question labels may represent information such as a category and an attribute of the questions, and by selecting an appropriate label for the questions, subsequent question queries or searches may be facilitated.

Referring to FIG. 4, an interface diagram of a creation interface is illustrated. The creation interface includes a question editing column and a label selection area. The question editing column can be used for a user to edit a question, and the user can edit a question mark as the end of the question editing. Optionally, the user may also supplement the detailed description of the question posed, e.g., motivation, context, etc. of the question, to provide additional reference information to other users replying to the question. Alternatively, the user may insert other types of information than text types, such as pictures, in the question editing column. The label selection area displays question labels which can be selected singly or in multiple modes for users, wherein the question labels can be custom labels or default labels provided for question and answer communities.

Further, when the user edits the question and selects the corresponding question label, a question control in the creation interface can be triggered (e.g. clicked, etc.), then the terminal can respond to the triggering operation of the user on the question control, send the question input by the user to the server side of the question-answering community to perform validity check, if the question passes the validity check, the server side can issue the question, and correspondingly, the user can display the issued question (i.e. the issue question in the application) on the interface of the client side of the question-answering community.

Referring to fig. 5, fig. 5 shows an interface schematic of a question-answering interface. The question-answering interface comprises a question display area, a question-answering communication area and a question-answering area, wherein the question-answering area comprises a first control used for triggering automatic answer to a target question, and optionally, the question-answering area also comprises an answer editing control used for editing and answering by a user, a question sharing control used for sharing and publishing the question by the user, and the like. Wherein the question display area is used for displaying the issued questions,

in some embodiments, the question display area may also display user information of the questioner corresponding to the issued question, such as a user nickname, a user avatar, etc.; further, the question display area may display a time of issuing the question, a place of issuing, a question label of issuing the question, and the like. The question-answer communication area is used for showing answers corresponding to the issued questions, and in the present application, for convenience of distinction, the answer to the current issued question is referred to as a target answer. Alternatively, the question and answer communication area may also be used to display an additional question or the like for the target answer.

In other embodiments, an entry into the question-and-answer interface (referred to as a first entry) may be provided in the answer preview interface for the issued question, upon which the question-and-answer interface may be displayed in response to a triggering operation to the first entry. The answer preview interface of the posting questions is used for displaying one or more answers to the posting questions, and when the content of the answer to the posting questions is long, the answer can be displayed in a folding manner in the answer preview interface of the posting questions. In this embodiment, the question-answer interface may include a question display area and an answer area, the answer area including the first control, and answers to the published questions may not need to be presented in the question-answer interface.

Optionally, referring to fig. 6, an interface schematic diagram of another question-answer interface is illustrated. As shown in fig. 6, the question-answer area of the question-answer interface displays two answer subregions: a first answer subregion and a second answer subregion, one for displaying one answer (i.e. target answer) to the question being issued, optionally the answer subregion may also display the answer time and the answer place. For example, the target answers of user a to the posted questions at the first answer time and first answer location are displayed in the first answer sub-zone.

Optionally, referring to fig. 7, an interface schematic diagram of yet another question-answer interface is illustrated. As shown in fig. 7, the first answer subarea of the question-answer interaction area of the question-answer interface displays a question for the target answer of the user a, which may be a question made by the user B for the target answer of the user a. It should be noted that, in the embodiment of the present application, the creation interface, the question-answer interface and the application component in each interface are all implemented based on a computing programming technology, and a virtual interface or a virtual component (which may also be understood as a control) with a visual application function is convenient for the user to perform man-machine interaction with the computer device. The display form and interaction mode of the virtual interface or the virtual component can be individually designed according to the interaction scene, and the method is not limited herein. For example, referring to fig. 8, fig. 8 shows an interface diagram of yet another problem interface. As shown in fig. 8, the question interface is a web page interface displayed on a browser.

Step 120: responding to the triggering operation of the first control to acquire the reply information of the target problem; the reply information is obtained by carrying out semantic coding and decoding based on the target problem.

The problem discussion is not time-efficient and even cannot obtain accurate and proper answers due to the fact that a period for obtaining a satisfactory answer is long after a user asks in a question-answering community. Therefore, the application provides automatic answers based on artificial intelligence for users so as to improve timeliness and accuracy of the answers to the questions. In the embodiment of the application, the first control on the question interface is used for triggering automatic reply to the target question or inquiring and commenting on the target answer.

The triggering operation may be operations such as clicking, touching, double clicking, long pressing, etc. triggered by the first control, which is not limited herein. The reply information is obtained by performing semantic coding and decoding based on the target problem, and specifically, if the target problem is a release problem, the reply information can be obtained by performing semantic coding and decoding on the release problem; if the target question is a challenge question for the target answer, the reply message may be obtained by performing semantic encoding and decoding on the issue question, the target answer of the issue question, and the challenge question.

In some embodiments, the user may trigger the first control on the question-answer interface, and further, in response to a triggering operation on the first control, the question-answer community client may perform semantic encoding and re-decoding based on the target question to generate reply information of the target question. Optionally, the client of the question-answer community may also respond to the triggering operation of the first control, send a reply request to the server of the question-answer community, respond to the service request, perform semantic encoding based on the target problem and then decode, obtain reply information of the target problem, and send the reply information of the target problem to the client, so that the client obtains the reply information of the target problem.

Referring to fig. 9, an exemplary diagram of a control response is shown in fig. 9, for one embodiment. As shown in fig. 9, the user may click on the first answer subregion on the question-answer interface, and further, a comment editing column corresponding to the target answer is displayed in the first answer subregion, and other users may edit and comment or query the target answer of the user a in the comment editing column, and click on the release control to release. Other users can also click the first control directly, the terminal responds to the clicking operation of the first control, the semantic encoding and decoding can be carried out on the basis of the issued questions and the target answers to generate the reply information of the target questions, and the reply information can be the inquiry or comment of the target answers.

As yet another embodiment, referring to fig. 10, fig. 10 shows another control response example graph. As shown in fig. 10, the user may click on the question and answer question of the target answer of the first answer subregion on the question and answer interface, and further, a comment editing column corresponding to the question and answer question is displayed in the first answer subregion, and other users may edit the question and answer question of the user B in the comment editing column and click on the issue control to issue. Other users can also click the first control directly, click operation of the terminal responding to the first control is carried out, semantic encoding and decoding are carried out on the basis of the issued questions, the target answers and the pursuit questions of the target answers, and then reply information of the target questions is generated, wherein the reply information can be pursuit or comment of the target answers.

Step 130: and displaying the reply information in association with the target problem, and displaying a first mark aiming at the reply information, wherein the first mark is used for indicating that the reply information is automatically generated.

In some embodiments, the first mark may be a designated avatar, and/or a nickname is designated, and if the first mark is a designated avatar, the first mark corresponds to the avatar that is the replying party to which the reply message corresponds; if the first mark is a designated nickname, the first mark corresponds to the nickname of the replying party as the replying information.

In other embodiments, the first marker may also be a designated avatar and a designated nickname. The specified avatar may include, for example, an avatar of an "AI" character or an avatar of an "intelligent reply" character, the specified nickname may be, for example, "intelligent reply", "automatic reply", etc., to explicitly prompt the user, so that the user may intuitively know that the reply information is automatically generated according to the specified avatar, and/or the specified nickname.

Referring to fig. 11, fig. 11 is a schematic diagram of an interface for displaying reply information. As shown in fig. 11, the terminal may display a posted question in a question display area of a question-and-answer interface, and display reply information automatically generated for the posted question and a first mark for the reply information in a question-and-answer communication area below the question display area, the first mark being a designated avatar for indicating that the reply information is automatically generated, the designated avatar shown in fig. 11 being merely an exemplary example and not to be construed as limiting the trial range of the present application.

Referring to fig. 12, fig. 12 is a schematic diagram of another interface for displaying reply information. As shown in fig. 12, the terminal may display reply information automatically generated for the target answer below the target answer of the user a to the posted question, and display a first mark for the reply information. As in fig. 11, the first mark is a specified header for indicating that the reply message is automatically generated.

Referring to fig. 13, fig. 13 is a schematic diagram of another interface for displaying reply information. As shown in fig. 13, the terminal may display reply information automatically generated for the target answer below the challenge question of the target answer of the user B to the user a, and display a first mark for the reply information. As in fig. 11, the first mark is a specified header for indicating that the reply message is automatically generated.

In other embodiments, the first mark may also be a preset identifier, where the preset identifier is not specifically limited herein, for example, "#", "", and the like, so as to distinguish reply information actually obtained by the user's work answer from automatically generated reply information through the preset identifier.

In other embodiments, the automatically generated reply message may not be displayed back to the compound avatar and nickname, thereby distinguishing reply messages actually obtained by the user's work from automatically generated reply messages.

In other embodiments, the automatically generated reply message may also be displayed by designating a display style, where the designated display style is different from a default display style of the reply message actually obtained by the user's work answer, so that the user is facilitated to intuitively recognize the automatically generated reply message.

In this embodiment, a question-answer interface may be displayed, where the question-answer interface includes a target question and a first control, the target question is a question to be issued or a question to be answered by a target, and the target answer is an answer to the issue question, further, response to a triggering operation on the first control is performed, the answer information of the target question is obtained by performing semantic encoding based on the target question and then decoding, and the answer information is associated with the target question and displayed with respect to the answer information, and a first mark is displayed with respect to the answer information.

Referring to fig. 14, fig. 14 is a flowchart illustrating a community question-answering method according to another embodiment of the present application. In this embodiment, the community question-answering method may be performed by a terminal having at least functions of display, storage, calculation, and communication. As shown in fig. 14, the community question-answering method specifically includes the following steps:

step 210: displaying a question-answer interface, wherein the question-answer interface comprises a target question and a first control; the target question is a question of issue or a question of inquiry for a target answer, which is an answer for the issue.

As one implementation mode, before a question and answer interface is displayed, a user establishes a release question for own questions in a question and answer community, performs validity check on the release question, and releases the release question after determining that the release question passes the validity check.

Referring to fig. 15, fig. 15 shows a problem distribution flowchart. After the terminal logs in the question-answering community, the user can enter a creation interface, the terminal can acquire the content of the issued questions and the selected question labels input by the user in the creation interface, and the content of the questions can comprise texts, pictures, videos and the like. Further, the terminal can perform validity verification on the issue problem, and if the issue problem is legal, the issue problem and related data thereof can be sent to the server for storage, so that the server can execute subsequent tasks based on the issue problem. Further, the terminal may display the posted questions on a question and answer interface.

Optionally, referring to fig. 16, fig. 16 shows a flow chart of a validity check. As shown in fig. 16, the terminal may display a question content input box and a question label selection menu at the creation interface, and the user may click on the submit control after completing the input of the question content and the question label selection. And the terminal can respond to the click operation of the user on the submitting control, send the issuing problem submitted by the user and the problem label to the server, and the server performs validity check on the issuing problem, if the server determines that the issuing problem is legal, the terminal issues the issuing problem and can display a question-answering interface aiming at the issuing problem after the issuing problem is issued. If the server side judges that the content is illegal, an error instruction is returned, the terminal receives the error instruction and displays error prompt information on the creation interface, and the error prompt information can be input again for prompting the user that the input content is illegal.

Specifically, step 210 may refer to the content of step 110 in the foregoing embodiment, which is not described herein.

Step 220: and responding to the triggering operation of the first control, and carrying out semantic coding based on the target problem to obtain target semantic coding characteristics.

As an embodiment, prior to semantic encoding, a text pre-process may be performed on the target question, where the text pre-process is used to clean text content in the target question, including removing irrelevant content, removing stop words, correcting spelling errors, word segmentation, etc., to improve the accuracy of the semantic analysis.

For example, remove extraneous content: advertisements, web page tags, messy codes and the like in the text content are removed. Removing stop words: words common in text content but with low semantic contribution are deleted. Correcting spelling errors: and correcting wrongly written characters in the text, and improving the quality and reliability of data. Stem extraction or morphological reduction: vocabulary is unified into its basic form by stem extraction (Stemming) or morphological reduction (Lemmatization).

Further, word segmentation (Token) may be performed on the text content, where the text content is segmented into individual words, characters, symbols, etc., to obtain a segmented word sequence, and these segmented elements are called tokens (Token).

Referring to fig. 17, fig. 17 shows a flow chart of text preprocessing. As shown in fig. 17, removing irrelevant content, word segmentation processing, lowercase, punctuation removal, stop word removal, spelling correction (i.e., correcting misspellings), and stem extraction or morphological restoration may be sequentially performed on the text content in the target question to obtain a word segmentation sequence.

Further, the word segmentation sequence can be subjected to semantic coding, and target semantic coding features are obtained. Specifically, word embedding may be performed on the Word segmentation sequence, that is, the middle Word element of the Word segmentation sequence is converted into a Word segmentation vector with a fixed size, and Word embedding may be performed by using models such as Word2Vec, gloVe, fastText, and the like, which are not limited herein.

Referring to fig. 18, fig. 18 illustrates a flowchart of semantic coding, where, as shown in fig. 18, the semantic coding may specifically include position coding (Positional Encoding) on a word vector to help understand a positional relationship of a word element in a sentence, for example, a position of each word element may be labeled by using a transducer model, so as to obtain a corresponding position code.

Further, a weight can be distributed to each word element according to the relation among the word elements based on a Self-attention mechanism (Self-Attention Mechanism), so that the contextual information of the word elements in the sentence is learned, long-distance dependence among the word elements is captured, and the whole sentence is better understood.

Further, after position coding and weight distribution are carried out on the segmented word vectors, multi-head attention and hierarchical structure (Multi-Head Attention and Layered Architecture) can be utilized to capture information of different segmented words in different semantic spaces, aggregation operation is carried out, semantic features of the segmented words are obtained through learning on different abstract levels based on a Multi-layer structure, and then target semantic coding features corresponding to target problems are obtained.

In some embodiments, if the target issue is a publishing issue, in step 220, the publishing issue may be semantically encoded, so as to obtain the target semantic encoding feature of the publishing issue. Correspondingly, in fig. 17, the text content is the text corresponding to the issue problem; the word segmentation sequences in fig. 17 and 18 are word segmentation sequences corresponding to the issue problem.

If the target question is an inquiry question for the target answer, in step 220, the issued question, the target answer and the inquiry question may be combined to obtain combined information, and then the combined information may be semantically encoded to obtain a target semantic encoding feature for the combined information. Correspondingly, in fig. 17, the text content is a text corresponding to the combination information; the word segmentation sequences in fig. 17 and 18 are word segmentation sequences corresponding to the combination information.

In some embodiments, the target question may include multi-modal content. For example, the question content may include pictures or videos in addition to text. Thus, the multi-mode content of the target problem can be semantically encoded in response to the triggering operation of the first control, and the target semantic encoding characteristics corresponding to the target problem are obtained.

Illustratively, the target question includes text content and image content. Performing multi-mode joint characterization learning on the text content and the image content to obtain joint feature representation of the text content and the image content, performing semantic coding on the joint feature representation to obtain target semantic coding features, further determining a target decoding strategy based on target response requirements determined on the target semantic coding features, and performing decoding processing on the target semantic coding features based on the target decoding strategy to obtain reply information of the target problem.

Optionally, in order to avoid the influence of the picture content in the target problem on the semantic coding of the text content, when multi-modal characterization learning is performed on the text content and the image content, corresponding text feature representation and image feature representation can be obtained on the text content and the image content respectively, the semantic similarity of the text content and the image content is determined based on the text feature representation and the image feature representation, and if the semantics of the text content and the image content are similar or identical, multi-modal joint characterization learning is performed on the text content and the image content. Otherwise, only carrying out semantic coding based on the text content of the target problem to obtain target semantic coding features.

Therefore, under the condition that the target problem comprises multi-mode content, content information of multiple modes can be combined for semantic coding, and the cooperative relationship among different modes of content is captured, so that reply information is generated more accurately. In addition, the problem of sparse data can be relieved by adding the contents of other modes such as pictures or videos on the text contents of the target problem, so that the semantic coding of the target problem has comprehensive feature learning.

Step 230: and determining the target response requirement of the target problem based on the target semantic coding features.

It is considered that in an actual application scenario, a user may have different response requirements for different types of questions. For example, an openness problem with creatives requires giving multiple replies. While problems with technical expertise resolution require accurate high quality replies. Therefore, the response requirement of the target problem, namely the target response requirement, can be determined, and then the decoding processing is carried out by using the adaptive decoding strategy according to different response requirements, so that the response information of the target problem is obtained.

As one embodiment, the target response requirements of the target problem may be accurately determined by quantifying the target response requirements, that is, classifying the response requirements based on the target semantic coding features. For example, the objective semantic coding feature may be classified by a classifier according to the answer requirements, the classification result is taken as the answer requirement, and the classification result may be the grade corresponding to different answer requirements.

In some embodiments, the answer demand classification may be performed on multiple answer demand dimensions, obtaining classification results on each answer demand dimension, such as a response speed dimension, an accuracy requirement dimension, an answer diversity dimension, and so on. Correspondingly, the classification result may include a response speed level in a response speed dimension, an accuracy requirement level in an accuracy requirement dimension, a diversity level in an answer diversity dimension, and the like.

In the scenario of performing answer demand classification on multiple answer demand dimensions, classifiers for each answer demand dimension may be trained separately, and then, the answer demand classification is performed on the target semantic coding feature under the answer demand dimension by using the classifier corresponding to the answer demand dimension. Wherein the response speed level is used to represent the efficiency of answering the target question. The accuracy requirement level is used to represent the accuracy requirement of the reply to the target problem. The diversity level is used to represent the diversity requirement of the reply to the target question. Alternatively, the type of the answer requirement and the number of dimensions of the answer requirement may be set according to a specific application scenario, which is not limited herein.

Of course, in other embodiments, the answer demand classification may also be performed based on the target semantic coding feature in one answer demand dimension.

Step 240: and determining a target decoding strategy corresponding to the target response requirement.

In the embodiment of the application, the target decoding strategy refers to a strategy for decoding target semantic coding features, which is determined according to target response requirements. In some embodiments, a correspondence between the response requirements and the decoding policies may be preset, after which, in step 240, the decoding policies corresponding to the target response requirements are correspondingly treated as target decoding policies.

Alternatively, the decoding strategies may include a Greedy Search decoding strategy (Greedy Search), a bundle Search decoding strategy (Beam Search), and a Sampling decoding strategy (Sampling).

The greedy search decoding strategy can be used for the problem with high requirement on response speed of the answer, for example, the problem of code running error reporting can be adopted. The bundle search decoding strategy may be used for questions with high accuracy requirements for the answer, e.g., questions with respect to interpretation of terms of art. The sample decoding strategy can be used for questions with high demands on the diversity of the answers, for example, questions that are creative in need of the opinion.

In one embodiment, if the target answer demand indicates that at least one of the response speed level is higher than the first speed level and the accuracy demand level is higher than the first accuracy level is satisfied, determining the target decoding policy to be a greedy search decoding policy.

As another embodiment, if the target answer demand indicates that at least one of the response speed level is not higher than the first speed level and the accuracy demand level is not higher than the first accuracy level is satisfied, determining the target decoding policy as a bundle search decoding policy.

As yet another embodiment, if the target answer demand indicates that the answer diversity level is higher than the first diversity level, the target decoding strategy is determined to be a sampling decoding strategy.

It should be noted that the determination of the first speed level, the first accuracy level, and the first diversity may be summarized through a lot of experiments, alternatively, other types of response requirement indexes such as the first speed level, the first accuracy level, and the first diversity may be set according to specific application scenarios, which are not limited herein, for example, in other embodiments, the target decoding policy may be determined at least based on the determination of the computing resource utilization level as the target response requirement.

Step 250: and decoding the target semantic coding features according to a target decoding strategy to obtain reply information of the target problem.

In embodiments of the present application, each decoding strategy may use a specific type of decoder. For example, pre-trained decoders of different types and corresponding decoding strategies may be stored in a relational database. Alternatively, the decoder may be a transducer network architecture.

As an implementation manner, after determining the target decoding policy, a corresponding target decoder may be determined for the target decoding policy from the relational database, where the decoder may decode according to the target decoding policy, and further, the target decoder decodes the target semantic coding feature according to the target decoding policy to obtain reply information of the target problem. Optionally, filtering operation may be performed on multiple candidate replies output by the target decoder, and after duplicate replies or illegal replies are deleted, the optimal reply is selected as reply information of the target problem according to the target reply requirement.

Step 260: and displaying the reply information in association with the target problem, and displaying a first mark aiming at the reply information, wherein the first mark is used for indicating that the reply information is automatically generated.

Specifically, step 260 may refer to the content of step 130 in the foregoing embodiment, which is not described herein.

In this embodiment, in response to a triggering operation of the first control of the question-answering interface, semantic coding may be performed based on the target question, and the target semantic coding feature may be obtained. And determining target response requirements of the target questions based on the target semantic coding features. Further, determining a target decoding strategy corresponding to the target response requirement, decoding the target semantic coding feature according to the target decoding strategy to obtain response information of the target problem, and displaying the response information and the target problem in an associated mode. Therefore, the method and the device realize that corresponding reply information is obtained by decoding according to decoding strategies corresponding to the current reply demands aiming at different reply demands, and aiming at the fact that questions of users have different reply demands in an actual application scene, the adaptive decoding strategy can be determined to generate reply contents meeting the reply demands of each question, and the reply information can be generated in a targeted manner under the scenes of different reply demands.

Referring to fig. 19, fig. 19 is a flowchart illustrating a community question-answering method according to another embodiment of the present application. In this embodiment, the community question-answering method may be performed by a terminal having at least functions of display, storage, calculation, and communication. As shown in fig. 19, the community question-answering method specifically includes the following steps:

Step 310: displaying a question-answer interface, wherein the question-answer interface comprises a target question and a first control; the target question is a question of issue or a question of inquiry for a target answer, which is an answer for the issue.

Step 320: and responding to the triggering operation of the first control, and carrying out semantic coding based on the target problem to obtain target semantic coding characteristics.

Specifically, the contents of step 310 and step 320 in the foregoing embodiments may be referred to as step 210 and step 220, and will not be described herein.

Step 330: and decoding the target semantic coding features according to at least two decoding strategies to obtain at least two candidate reply messages of the target problem.

The target semantic coding features can be decoded by utilizing various decoding strategies to obtain a plurality of candidate reply messages, and reply messages of the target problem can be selected from the plurality of candidate reply messages according to different evaluation dimensions. In some embodiments, the at least two decoding strategies may be at least two of the greedy search decoding strategy, the bundle search decoding strategy, and the sample decoding strategy above. Of course, in other embodiments, the decoding strategy is not limited thereto.

As an implementation manner, the target semantic coding feature can be decoded by decoders corresponding to at least two decoding strategies to obtain corresponding candidate reply information. The step of obtaining the decoder may refer to the content of step 250 in the foregoing embodiment, which is not described herein. And decoding the target semantic coding features through at least two decoding strategies to obtain a plurality of candidate reply messages of the target problem, so that the reply messages are automatically generated by diversity, and the better reply messages can be conveniently selected from the diversity reply messages.

Referring to fig. 20, fig. 20 shows a flowchart of a decoding process. As shown in fig. 20, the target semantic coding feature may be decoded by decoders corresponding to different decoding strategies to obtain multiple candidate reply messages. Specifically, the greedy search decoding strategy, the bundle search decoding strategy and the sampling decoding strategy can respectively decode the target semantic coding feature to obtain respective corresponding candidate reply information, evaluate the candidate reply information to obtain an evaluation result of the candidate reply information, and then determine the target reply information from the candidate reply information according to the evaluation result.

Step 340: and evaluating each candidate reply message to obtain an evaluation result of each candidate reply message.

In some embodiments, each candidate reply message may be evaluated under a preset evaluation dimension. The preset evaluation dimension may be at least one of an accuracy evaluation dimension, a readability evaluation dimension, a conciseness evaluation dimension, a diversity evaluation dimension, and a consistency evaluation dimension.

The readability evaluation dimension is used for evaluating the reply information under the readability dimension; the conciseness evaluation dimension is used for evaluating the reply information under the conciseness dimension. The consistency evaluation dimension is used for evaluating the reply information under the consistency dimension of the reply phase relative to the problem, such as evaluating the consistency of the reply information relative to the problem in terms of logic, theme, content and the like. For example, logical consistency: the reply message needs to be logically consistent, without paradox or confusing implications. For example, when answering a question, each statement in the answer information should effectively support or deduce a conclusion that the entire answer is to be in compliance with logic principles.

Theme consistency: all information and discussions in the reply message should be related to the subject matter of the question, avoiding the occurrence of irrelevant detailed information in the reply, or jumping to topics that are irrelevant to the question. Content consistency: all facts and data in the reply message are accurate and consistent. For example, in interpreting a concept, it is ensured that the data or facts of the reply references of the different parts do not contradict each other.

The evaluation result of the candidate reply message may refer to a score of the candidate reply message, and if the candidate reply message is evaluated in a plurality of evaluation dimensions, the evaluation result of the candidate reply message includes the score of the candidate reply message in each of the plurality of evaluation dimensions.

As an embodiment, each candidate reply message may be scored under at least two evaluation dimensions, and the score of each candidate reply message under each evaluation dimension is obtained, where the at least two evaluation dimensions include at least two of an accuracy evaluation dimension, a readability evaluation dimension, a conciseness evaluation dimension, and a diversity evaluation dimension.

Further, for each candidate reply message, the scores of the candidate reply message under a plurality of evaluation dimensions can be weighted according to the evaluation weights respectively corresponding to at least two evaluation dimensions, so as to obtain the evaluation result of the candidate reply message.

Step 350: and determining reply information of the target problem from at least two candidate reply information based on the evaluation result of each candidate reply information.

In some embodiments, the score of each candidate reply message may be determined according to the evaluation result of each candidate reply message, where the score is used to reflect the quality of the candidate reply message, that is, the higher the score, the higher the quality of the candidate reply message, and on this basis, the candidate reply message with the highest score may be used as the reply message of the target problem.

In other embodiments, if the evaluation result of the candidate reply information includes scores in multiple evaluation dimensions, the scores of the candidate reply information in the multiple evaluation dimensions may be weighted according to the weights corresponding to each evaluation dimension, so as to obtain a composite score of each candidate reply information, where the higher the composite score, the higher the quality of the candidate reply information. Then, the candidate reply information with the highest comprehensive score can be used as the reply information of the target problem. The weights for different evaluation dimensions may be the same or different, and may be set according to actual needs, and are not particularly limited herein.

Step 360: and displaying the reply information in association with the target problem, and displaying a first mark aiming at the reply information, wherein the first mark is used for indicating that the reply information is automatically generated.

Step 360 may refer to the content of step 130 in the foregoing embodiment, and will not be described herein.

In this embodiment, candidate reply information generated by multiple decoding strategies is evaluated in one or more evaluation dimensions, for example, from multiple evaluation dimensions such as accuracy, readability, consistency, conciseness and diversity, so that reply information with better quality can be screened out from multiple candidate reply information, thereby effectively ensuring the quality of the reply information.

Referring to fig. 21, fig. 21 is a schematic flow chart of a community question-answering method according to still another embodiment of the present application. In this embodiment, the community question-answering method may be executed by a terminal device having at least functions of display, storage, calculation, and communication. As shown in fig. 20, the community question-answering method specifically includes the following steps:

step 410: displaying a question-answer interface, wherein the question-answer interface comprises a target question and a first control; the target question is an pursuit question for a target answer, which is an answer to a posting question.

Specifically, step 410 may refer to the content of step 210 in the foregoing embodiment, which is not described herein.

Step 420: and responding to the triggering operation of the first control, and combining the issued question, the target answer and the inquiry question aiming at the target answer to obtain combined information.

In an actual application scene, considering that a user may possibly put forward an inquiry question for a target answer, in order to improve the accuracy and timeliness of the reply to the inquiry question, corresponding reply information can be automatically generated for the inquiry question. Specifically, the generation of the reply information can be performed based on the combined information of the issued question, the target answer and the inquiry question aiming at the target answer, and the context relationship in the combined information can be better utilized to accurately perform semantic coding through the combined information, so that more information is provided for the subsequent generation of the reply information.

In some embodiments, the issue question, the target answer, and the challenge question for the target answer may be stitched to obtain the combined information.

As an embodiment, referring to fig. 8, the terminal may combine the issued question, the target answer, and the target answer query in response to a click operation performed by the user on the target answer query in the first answer subregion on the question and answer interface, to obtain a corresponding combined text, that is, combined information.

Optionally, considering that the target question is an overtime question for the target answer, the issuing question is only used for providing context information for the overtime question, if the issuing question includes non-text mode content such as images or videos, the text content in the issuing question can be spliced with the target answer and the overtime question for the target answer to obtain combined information, so that the calculation complexity of semantic coding of the combined information can be reduced, and the accuracy of the target semantic coding feature can be improved.

Step 430: and carrying out semantic coding on the combined information to obtain target semantic coding features.

As an implementation manner, the terminal may perform semantic coding on the combined text to obtain the target semantic coding feature, and specifically, step 430 may refer to the content of step 220 in the foregoing embodiment, which is not described herein again.

Step 440: reply information is generated based on the target semantic coding features.

As one implementation mode, the terminal can determine the target response requirement of the target problem based on the target semantic coding features, and decode the target semantic coding features through a target decoding strategy determined by the target response requirement to obtain the response information of the target problem. The terminal can also decode the target semantic coding feature according to at least two decoding strategies to obtain at least two candidate reply messages of the target problem, and determine the reply message of the target problem from the at least two candidate reply messages based on the evaluation result of each candidate reply message. Specifically, step 430 may refer to the contents of steps 230 to 250 and steps 330 to 350 in the foregoing embodiments, and will not be described herein.

Step 450: and displaying the reply information in association with the target problem, and displaying a first mark aiming at the reply information, wherein the first mark is used for indicating that the reply information is automatically generated.

Specifically, step 450 may refer to the content of step 130 in the foregoing embodiment, which is not described herein.

In some embodiments, after step 110, the method further includes the following steps 460-470, described in detail below:

Step 460: in response to the input operation, a first reply to the target question input is obtained.

Step 470: and responding to the issuing operation triggered by the first reply, and displaying the first reply and the target problem in an associated mode.

The input operation may be an edit operation of a user in the question-answering community to answer the target question. The input operation may include text editing, image insertion, video insertion, and the like. The publishing operation may be an operation on the user-triggered question-answer interface for requesting a virtual control to publish the first reply, such as an operation to click on the virtual control.

After a user logs in the question and answer platform, the user can enter a question and answer interface, edit and answer the question by taking the target question as the issue question in an edit box of an answer area of the question and answer interface, and click the issue control when the edition is completed, so that the terminal can respond to the click operation of the issue control by the user, and the edit content of the user is displayed as a first answer in the question and answer exchange area of the question and answer interface.

In this embodiment, in the process of presenting a query to a target answer by a user of the question-answer community, the query questions of the issued question, the target answer and the target answer are combined and then semantically encoded to generate the reply information. Therefore, because some unclear questions are subjected to the following process, the reply information is more and more accurate, the questions are more and more clear, and the reply quality is improved.

Referring to fig. 22, fig. 22 is a timing chart illustrating a community question-answering method according to another embodiment of the present application. In this embodiment, the community question-answering method may be cooperatively executed by the server and the terminal. Illustratively, the Server and the terminal may form a Client-Server (C/S) system architecture, the Server may be a cloud Server deployed by a service provider, and the terminal may be a smart phone, a tablet computer, a desktop computer, a vehicle-mounted terminal, a learning terminal, or the like; the terminal can operate the client corresponding to the question-answer community. In fig. 22, steps 510 and 550 may be performed by the terminal, more specifically, by a client corresponding to the question-answer community operated by the terminal, and steps 520 to 540 are performed by the server. As shown in fig. 22, the community question-answering method specifically includes the following steps:

step 510: and sending a reply request to the server.

Step 520: receiving a reply request sent by the client, wherein the reply request is generated by the client in response to the triggering operation of a first control in a question-answer interface, the question-answer interface further comprises a target question, the target question is a release question or a query question aiming at a target answer, and the target answer is an answer aiming at the release question.

As one implementation, the terminal may generate a reply request in response to a triggering operation of the user on the first control in the question-answer interface, and send the reply request to the server. The reply request is used for requesting the server to automatically generate reply information for the target problem in the question-answering interface. Further, the server may receive a reply request sent by the client.

Step 530: and responding to the reply request, carrying out semantic coding and decoding based on the target problem, and obtaining reply information of the target problem.

Considering that if the first control is displayed on the question-answer interface of each client for the target problem, when the questioning user selects to automatically generate the answer for the target problem, other users trigger the first control, and the answer can be not automatically generated any more, thereby preventing repeated generation of the answer information.

In one embodiment, the server responds to the reply request to determine whether to generate reply information for the target question, and if so, generates a rejection instruction and sends the rejection instruction to the client to prompt the user that the answer has been automatically replied. Alternatively, the question interface with the first control may be displayed for the asking user of the target question, and the question interface without the first control may be displayed for the non-asking user of the target question.

Step 540: and sending the reply information of the target problem to the client so that the client can correlate and display the reply information with the target problem, and displaying a first mark aiming at the reply information, wherein the first mark is used for indicating that the reply information is automatically generated.

As one implementation, the server may perform semantic encoding and decoding based on the target problem in response to the received reply request, to obtain reply information of the target problem. Specifically, referring to fig. 23, fig. 23 shows a flowchart of reply message generation. The server can perform text preprocessing on the target problem, input the text after the text preprocessing into the language model for word embedding to obtain a corresponding word segmentation sequence, further input the word segmentation sequence into the semantic coding model for semantic coding to obtain target semantic coding features, and further decode the target semantic coding features to obtain reply information of the target problem.

Optionally, the server may determine a target response requirement of the target problem based on the target semantic coding feature, and decode the target semantic coding feature according to a target decoding policy determined by the target response requirement to obtain reply information of the target problem. For specific implementation details, reference may be made to the description of the corresponding procedure in the corresponding embodiment of fig. 14, which is not repeated here.

Optionally, the server may perform decoding processing on the target semantic coding feature according to at least two decoding strategies to obtain at least two candidate reply messages of the target problem, and determine reply messages of the target problem from the at least two candidate reply messages based on an evaluation result of each candidate reply message. Specifically, step 540 may refer to the contents of steps 230 to 250 and 330 to 350 in the foregoing embodiments, and will not be described herein.

Step 550: and receiving reply information, carrying out associated display on the reply information and the target problem, and displaying a first mark aiming at the reply information.

Specifically, step 550 may refer to the content of step 130 in the foregoing embodiment, which is not described herein.

Referring to fig. 24, a block diagram of a community question-answering apparatus 600 according to an embodiment of the present application is shown. The community question-answering apparatus 600 includes:

a question and answer display module 610, configured to display a question and answer interface, where the question and answer interface includes a target question and a first control; the target question is a release question or an overt question aiming at a target answer, and the target answer is an answer aiming at the release question;

an information obtaining module 620, configured to obtain, in response to a triggering operation on the first control, reply information of the target problem, where the reply information is obtained by performing semantic encoding and then decoding based on the target problem;

And a reply display module 630, configured to display the reply message in association with the target question, and display a first flag for the reply message, where the first flag is used to indicate that the reply message is automatically generated.

In some embodiments, the information acquisition module 620 may include: the system comprises a first encoding unit, a demand determining unit, a strategy determining unit and an information generating unit. The first coding unit is used for responding to the triggering operation of the first control, carrying out semantic coding based on the target problem and obtaining target semantic coding characteristics; the demand determining unit is used for determining the target response demand of the target problem based on the target semantic coding features; the strategy determining unit is used for determining a target decoding strategy corresponding to the target response requirement; and the information generating unit is used for carrying out decoding processing on the target semantic coding features according to the target decoding strategy to obtain the reply information of the target problem.

In some embodiments, the policy determination unit may be specifically configured to: if the target answer demand indicates that at least one of a response speed level higher than a first speed level and an accuracy demand level higher than a first accuracy level is met, determining that the target decoding strategy is a greedy search decoding strategy; if the target response requirement indicates that at least one of the response speed level is not higher than the first speed level and the accuracy requirement level is not higher than the first accuracy level is met, determining that the target decoding strategy is a cluster search decoding strategy; and if the target response requirement indicates that the reply diversity level is higher than the first diversity level, determining that the target decoding strategy is a sampling decoding strategy.

In some embodiments, the demand determination unit may be specifically configured to: and classifying response demands based on the target semantic coding features, and determining target response demands of the target problems.

In some embodiments, the information acquisition module 620 may further include: the device comprises a second encoding unit, a decoding processing unit, a result evaluation unit and a reply confirmation unit. The second coding unit is used for responding to the triggering operation of the first control, carrying out semantic coding based on the target problem and obtaining target semantic coding characteristics; the decoding processing unit is used for decoding the target semantic coding features according to at least two decoding strategies to obtain at least two candidate reply messages of the target problem; the result evaluation unit is used for evaluating each candidate reply message to obtain an evaluation result of each candidate reply message; and the reply confirmation unit is used for determining reply information of the target problem from the at least two candidate reply information based on the evaluation result of each candidate reply information.

In some embodiments, the result evaluation unit may be specifically configured to: scoring each candidate reply message under at least two evaluation dimensions, and obtaining the score of each candidate reply message under each evaluation dimension; and for each candidate reply message, weighting the scores of the candidate reply message under a plurality of evaluation dimensions according to the evaluation weights respectively corresponding to at least two evaluation dimensions to obtain an evaluation result of the candidate reply message.

In some embodiments, the at least two assessment dimensions include at least two of an accuracy assessment dimension, a readability assessment dimension, a conciseness assessment dimension, and a diversity assessment dimension.

In some embodiments, the target question is an pursuit question for a target answer; the first encoding unit and the second encoding unit may be specifically configured to: responding to the triggering operation of the first control, and combining the issued questions, the target answers and the inquiry questions aiming at the target answers to obtain combined information; and carrying out semantic coding on the combined information to obtain the target semantic coding features.

In some embodiments, community question and answer device 600 may further include: an input module and a release module. The input module is used for responding to input operation and acquiring a first reply input aiming at the target problem; and the issuing module is used for responding to the issuing operation triggered by the first reply and carrying out associated display on the first reply and the target problem.

It will be clearly understood by those skilled in the art that, for convenience and brevity of description, the specific working process of the apparatus and modules described above may refer to the corresponding process in the foregoing method embodiment, which is not repeated herein.

In several embodiments provided by the present application, the coupling of the modules to each other may be electrical, mechanical, or other.

In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist alone physically, or two or more modules may be integrated into one module. The integrated modules may be implemented in hardware or in software functional modules.

Therefore, after the user issues the question by the question-answering community, the first control can be touched, so that the question-answering community can perform semantic coding and decoding on the question issued by the user to obtain the reply information, and the reply information is automatically displayed on the question-answering interface, thereby avoiding the question-answering user from passively waiting for the answer of other users to the question for a long time and improving the timeliness of the question reply.

Referring to fig. 25, a block diagram of a community question-answering apparatus 700 according to an embodiment of the present application is shown. The community question-answering device 700 is applied to a server, and the community question-answering device 700 includes:

a request receiving module 710, configured to receive a reply request sent by a client, where the reply request is generated by the client in response to a triggering operation on a first control in a question-answer interface, and the question-answer interface further includes a target question, where the target question is a issued question or an pursuit question for a target answer, and the target answer is an answer for the issued question;

The reply generation module 720 is configured to respond to the reply request, perform semantic encoding based on the target problem, and then decode the target problem, so as to obtain reply information of the target problem;

and a reply sending module 730, configured to send reply information of the target problem to the client, so that the client displays the reply information in association with the target problem, and displays a first mark for the reply information, where the first mark is used to indicate that the reply information is automatically generated.

It will be clearly understood by those skilled in the art that, for convenience and brevity of description, the specific working process of the apparatus and modules described above may refer to the corresponding process in the foregoing method embodiment, which is not repeated herein.

In several embodiments provided by the present application, the coupling of the modules to each other may be electrical, mechanical, or other.

In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist alone physically, or two or more modules may be integrated into one module. The integrated modules may be implemented in hardware or in software functional modules.

As shown in fig. 26, an embodiment of the present application further provides a computer apparatus 800, where the computer apparatus 800 includes a processor 810, a memory 820, a power source 830, and an input unit 840, and the memory 820 stores a computer program, and when the computer program is called by the processor 810, the computer program can implement the various method steps provided in the above embodiments. It will be appreciated by those skilled in the art that the structure of the computer device shown in the drawings does not constitute a limitation of the computer device, and may include more or less components than those illustrated, or may combine certain components, or may be arranged in different components. Wherein:

processor 810 may include one or more processing cores. The processor 810 connects various parts within the overall battery management system using various interfaces and lines, and overall controls the computer device by executing or executing instructions, programs, instruction sets, or program sets stored in the memory 820, invoking data stored in the memory 820, performing various functions of the battery management system and processing data, and executing various functions of the computer device and processing data. Alternatively, the processor 810 may be implemented in hardware in at least one of digital signal processing (Digital Signal Processing, DSP), field programmable gate array (Field-Programmable Gate Array, FPGA), programmable logic array (Programmable Logic Array, PLA). The processor 810 may integrate one or a combination of several of a central processing unit 810 (Central Processing Unit, CPU), an image processor 810 (Graphics Processing Unit, GPU), and a modem, etc. The CPU mainly processes an operating system, a user interface, an application program and the like; the GPU is used for being responsible for rendering and drawing of display content; the modem is used to handle wireless communications. It will be appreciated that the modem may not be integrated into the processor 810 and may be implemented solely by a single communication chip.

The Memory 820 may include a random access Memory 820 (Random Access Memory, RAM) or a Read-Only Memory 820 (Read-Only Memory). Memory 820 may be used to store instructions, programs, sets of instructions, or program sets. The memory 820 may include a stored program area and a stored data area, wherein the stored program area may store instructions for implementing an operating system, instructions for implementing at least one function (such as a touch function, a sound playing function, an image playing function, etc.), instructions for implementing the various method embodiments described above, and the like. The storage data area may also store data created by the computer device in use, such as phonebook and audio video data, and the like. Accordingly, memory 820 may also include a memory controller to provide access to memory 820 by processor 810.

The power supply 830 may be logically connected to the processor 810 through a power management system, so as to perform functions of managing charging, discharging, and power consumption management through the power management system. The power supply 830 may also include one or more of any components, such as a direct current or alternating current power supply, a recharging system, a power failure detection circuit, a power converter or inverter, a power status indicator, and the like.

An input unit 840, the input unit 840 being operable to receive input numeric or character information and to generate keyboard, mouse, joystick, optical or trackball signal inputs related to user settings and function control.

Although not shown, the computer device 800 may further include a display unit or the like, which is not described herein. In particular, in this embodiment, the processor 810 in the computer device loads executable files corresponding to the processes of one or more computer programs into the memory 820 according to the following instructions, and the processor 810 executes the program files stored in the memory 820, such as phonebook and audio video data, so as to implement the various method steps provided in the foregoing embodiment.

As shown in fig. 27, an embodiment of the present application further provides a computer readable storage medium 900, where the computer readable storage medium 900 stores a computer program 910, and the computer program 910 may be called by a processor to perform various method steps provided by the embodiment of the present application.

The computer readable storage medium may be an electronic memory such as a flash memory, an EEPROM (electrically erasable programmable read only memory), an EPROM, a hard disk, or a ROM. Optionally, the computer readable storage medium comprises a Non-volatile computer readable storage medium (Non-Transitory Computer-Readable Storage Medium). The computer readable storage medium 900 has storage space for a computer program that performs any of the method steps in the embodiments described above. These computer programs may be read from or written to one or more computer program products. The computer program can be compressed in a suitable form.

According to one aspect of the present application, there is provided a computer program product comprising a computer program stored in a computer readable storage medium. The processor of the computer device reads the computer program from the computer-readable storage medium, and the processor executes the computer program so that the computer device performs the various method steps provided by the above embodiments.

Although the present application has been described in terms of the preferred embodiments, it should be understood that the present application is not limited to the specific embodiments, but is capable of numerous modifications and equivalents, and alternative embodiments and modifications of the embodiments described above, without departing from the spirit and scope of the present application.

Claims (15)

1. A community question-answering method, the method comprising:

displaying a question-answer interface, wherein the question-answer interface comprises a target problem and a first control; the target question is a release question or an overt question aiming at a target answer, and the target answer is an answer aiming at the release question;

Responding to the triggering operation of the first control, and acquiring reply information of the target problem, wherein the reply information is obtained by carrying out semantic coding and decoding on the basis of the target problem;

and displaying the reply information in association with the target problem, and displaying a first mark aiming at the reply information, wherein the first mark is used for indicating that the reply information is automatically generated.

2. The method of claim 1, wherein the obtaining reply information to the target question in response to the triggering operation of the first control comprises:

responding to the triggering operation of the first control, and carrying out semantic coding based on the target problem to obtain target semantic coding characteristics;

determining a target response requirement of the target problem based on the target semantic coding features;

determining a target decoding strategy corresponding to the target answering requirement;

and decoding the target semantic coding features according to the target decoding strategy to obtain the reply information of the target problem.

3. The method of claim 2, wherein determining the target decoding policy corresponding to the target answer demand comprises:

If the target answer demand indicates that at least one of a response speed level higher than a first speed level and an accuracy demand level higher than a first accuracy level is met, determining that the target decoding strategy is a greedy search decoding strategy;

if the target response requirement indicates that at least one of the response speed level is not higher than the first speed level and the accuracy requirement level is not higher than the first accuracy level is met, determining that the target decoding strategy is a cluster search decoding strategy;

and if the target response requirement indicates that the reply diversity level is higher than the first diversity level, determining that the target decoding strategy is a sampling decoding strategy.

4. The method of claim 2, wherein the determining the target answer requirement for the target question based on the target semantic coding feature comprises:

and classifying response demands based on the target semantic coding features, and determining target response demands of the target problems.

5. The method of claim 1, wherein the obtaining reply information to the target question in response to the triggering operation of the first control comprises:

responding to the triggering operation of the first control, and carrying out semantic coding based on the target problem to obtain target semantic coding characteristics;

Decoding the target semantic coding features according to at least two decoding strategies to obtain at least two candidate reply messages of the target problem;

evaluating each candidate reply message to obtain an evaluation result of each candidate reply message;

and determining reply information of the target problem from the at least two candidate reply information based on the evaluation result of each candidate reply information.

6. The method of claim 5, wherein the evaluating each candidate reply message to obtain an evaluation result of each candidate reply message comprises:

scoring each candidate reply message under at least two evaluation dimensions, and obtaining the score of each candidate reply message under each evaluation dimension;

and for each candidate reply message, weighting the scores of the candidate reply message under a plurality of evaluation dimensions according to the evaluation weights respectively corresponding to at least two evaluation dimensions to obtain an evaluation result of the candidate reply message.

7. The method of claim 6, wherein the at least two assessment dimensions comprise at least two of an accuracy assessment dimension, a readability assessment dimension, a conciseness assessment dimension, and a diversity assessment dimension.

8. The method of claim 2 or 5, wherein the target question is a challenge question for a target answer;

the responding to the triggering operation of the first control performs semantic coding based on the target problem to obtain target semantic coding characteristics, and the method comprises the following steps:

responding to the triggering operation of the first control, and combining the issued questions, the target answers and the inquiry questions aiming at the target answers to obtain combined information;

and carrying out semantic coding on the combined information to obtain the target semantic coding features.

9. The method of claim 1, wherein after the displaying the question-answer interface, the method further comprises:

responding to input operation, and acquiring a first reply input aiming at the target problem;

and responding to the issuing operation triggered by the first reply, and carrying out associated display on the first reply and the target problem.

10. A community question-answering method, characterized by being applied to a server, the method comprising:

receiving a reply request sent by a client, wherein the reply request is generated by the client in response to a triggering operation of a first control in a question-answer interface, and the question-answer interface further comprises a target question, wherein the target question is a release question or an overtime question aiming at a target answer, and the target answer is an answer aiming at the release question;

Responding to the reply request, carrying out semantic coding and decoding based on the target problem, and obtaining reply information of the target problem;

and sending reply information of the target problem to the client so that the client can correlate and display the reply information with the target problem, and displaying a first mark aiming at the reply information, wherein the first mark is used for indicating that the reply information is automatically generated.

11. A community question-answering apparatus, the apparatus comprising:

the question and answer display module is used for displaying a question and answer interface, and the question and answer interface comprises a target question and a first control; the target question is a release question or an overt question aiming at a target answer, and the target answer is an answer aiming at the release question;

the information acquisition module is used for responding to the triggering operation of the first control to acquire the reply information of the target problem, wherein the reply information is obtained by carrying out semantic coding and decoding on the basis of the target problem;

and the reply display module is used for carrying out associated display on the reply information and the target problem, and displaying a first mark aiming at the reply information, wherein the first mark is used for indicating that the reply information is automatically generated.

12. A community question-answering device, characterized in that it is applied to a server, said device comprising:

the request receiving module is used for receiving a reply request sent by the client, wherein the reply request is generated by the client in response to the triggering operation of a first control in a question-answer interface, the question-answer interface further comprises a target question, the target question is a release question or an overtime question aiming at a target answer, and the target answer is an answer aiming at the release question;

the reply generation module is used for responding to the reply request, carrying out semantic coding and decoding again based on the target problem and obtaining reply information of the target problem;

the reply sending module is used for sending reply information of the target problem to the client so that the client can correlate and display the reply information with the target problem, and a first mark is displayed for the reply information, wherein the first mark is used for indicating that the reply information is automatically generated.

13. A computer readable storage medium having stored therein program code which is callable by a processor to perform the method according to any one of claims 1 to 10.

14. A computer device, comprising:

a memory;

one or more processors coupled with the memory;

one or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the one or more processors, the one or more applications configured to perform the method of any of claims 1-10.

15. A computer program product or computer program, characterized in that the computer program product or computer program comprises computer instructions stored in a storage medium; a processor of a computer device reads the computer instructions from a storage medium, the processor executing the computer instructions, causing the computer to perform the method of any of claims 1-10.