CN115687766A - Case retrieval push method and device, storage medium and electronic equipment - Google Patents

Abstract

The disclosure provides a case retrieval pushing method and related equipment. The method comprises the following steps: acquiring text data to be retrieved of a case to be retrieved; processing the text data to be retrieved through the case element recognition model to obtain target element characteristics of each element of the case to be retrieved; acquiring historical element characteristics of each element of a plurality of historical cases, and acquiring the current pushed times; updating a retrieval model based on the pushed times, and comparing the target element characteristics of the case to be retrieved with the historical element characteristics of each historical case through the updated retrieval model so as to determine an initial pushed case sequence from the plurality of historical cases; and determining a target pushing case sequence of the case to be retrieved according to the initial pushing case sequence so as to push and display the historical cases in the target pushing case sequence. The method can realize implicit feedback of the result generated by the retrieval model, so that the case retrieval pushed result is more accurate, and the personalized retrieval requirement of a user can be met.

Description

Case retrieval push method and device, storage medium and electronic equipment

technical field

The present disclosure relates to the field of computer technology, and in particular to a case retrieval and push method and device, a storage medium, and electronic equipment.

Background technique

With the development of network technology, most of the judgment documents of historical cases can be publicly searched and retrieved on the Internet, and the search of similar cases can ensure the uniform judgment standards of similar cases.

In the related technology, usually relevant personnel search for historical cases similar to the case to be retrieved from the Internet based on experience, which is low in efficiency and poor in accuracy.

It should be noted that the information disclosed in the above background section is only for enhancing the understanding of the background of the present disclosure, and therefore may include information that does not constitute the prior art known to those of ordinary skill in the art.

Contents of the invention

The purpose of the present disclosure is to provide a case retrieval push method, device, electronic equipment and storage medium to improve the efficiency and accuracy of similar case retrieval results.

Other features and advantages of the present disclosure will become apparent from the following detailed description, or in part, be learned by practice of the present disclosure.

According to one aspect of the present disclosure, there is provided a case retrieval push method, including: obtaining the text data to be retrieved of the case to be retrieved; processing the text data to be retrieved through a case element recognition model to obtain the target element features of each element of the case to be retrieved; Obtain the historical element characteristics of each element of multiple historical cases, and obtain the current number of pushes; update the retrieval model based on the number of pushes, and use the updated retrieval model to use the target element characteristics of the case to be retrieved and the historical element characteristics of each historical case Perform comparison processing to determine the initial push case sequence from multiple historical cases; determine the target push case sequence of the cases to be retrieved according to the initial push case sequence, so as to push and display the historical cases in the target push case sequence.

In one embodiment of the present disclosure, the retrieval model is updated based on the number of push times, and the updated retrieval model is used to compare the target element features of the case to be retrieved with the historical element features of each historical case to determine from multiple historical cases The initial pushed case sequence includes: updating the current weight value of each element in the retrieval model based on the number of push times; calculating the similarity between the target element features and each historical element feature based on the current weight value, and obtaining the relationship between the case to be retrieved and each historical case. The similarity between multiple historical cases is sorted according to the similarity to obtain the sorting result of multiple historical cases; the historical cases of the previous preset number in the sorting result are determined as the initial push case sequence.

In an embodiment of the present disclosure, updating the current weight value of each element in the retrieval model based on the number of push times includes: if the number of push times is equal to 0, then determining the current weight value corresponding to each element as a preset weight value; if If the number of push times is greater than 0, get the user behavior feedback data in the latest push, and get the historical weight value of each element used in the retrieval model in the latest push; update the historical weight value according to the user behavior feedback data to get The current weight value for each feature.

In one embodiment of the present disclosure, obtaining the user behavior feedback data in the latest push includes: obtaining the user's browsing time for each target historical case in the target push case sequence of the latest push, and obtaining a preset duration threshold; according to The duration threshold converts the browsing time into a behavior feedback value; the user behavior feedback data of each target historical case is determined according to the behavior feedback value of each target historical case.

In an embodiment of the present disclosure, the historical weight value is updated according to the user behavior feedback data to obtain the current weight value of each element, including: obtaining the relationship between each target historical case and the case to be retrieved in the target push case sequence pushed last time The similarity between them is used to form the environmental data, and the historical weight is used as the action data; the environmental data, action data and user behavior feedback data are merged to obtain the input data; the input data is input into the trained deep reinforcement learning model , the output gets the current weight value of each element.

In one embodiment of the present disclosure, the multiple historical cases include the first case; the elements include the first element; wherein, according to the current weight value, the similarity calculation is performed on the characteristics of the target element and the characteristics of each historical element, and the case to be retrieved and each historical element are obtained. The similarity between cases includes: using a similarity algorithm to calculate the similarity between the target element characteristics of the first element of the case to be retrieved and the historical element characteristics of the first element of the first case, and obtain the case to be retrieved and the first case The element similarity between the first element and the element similarity between the case to be retrieved and the first case are obtained; the weighted calculation is performed according to the current weight value of each element and the corresponding element similarity to obtain the element similarity between the case to be retrieved and the first case. The similarity between the case and the first case.

In one embodiment of the present disclosure, determining the target push case sequence of the case to be retrieved according to the initial push case sequence includes: if the number of push cases is equal to 0, then use the initial push case sequence as the target push case sequence; if the push count is greater than 0 , then obtain the latest pushed historical push case sequence, and determine the target push case sequence according to the historical push case sequence and the initial push case sequence.

In an embodiment of the present disclosure, the case retrieval push method further includes: acquiring historical text data of multiple historical cases; processing each historical text data through a case element recognition model to obtain the historical element characteristics of each historical case; The text data and its historical element features are correspondingly stored in the case database; and in the method, acquiring the historical element features of multiple historical cases includes: acquiring the historical element features of multiple historical cases from the case database.

In an embodiment of the present disclosure, the element features include at least one of the following: legal relationship features, core fact features, and evidentiary circumstances features.

According to another aspect of the present disclosure, a case retrieval push device is provided, including:

The acquisition module is used to obtain the text data to be retrieved of the case to be retrieved; the processing module is used to process the text data to be retrieved through the case element recognition model to obtain the target element characteristics of each element of the case to be retrieved; the acquisition module is also used to obtain multiple The historical element characteristics of each element of a historical case, and obtain the current number of pushes; the processing module is also used to update the retrieval model based on the number of pushes, and use the updated retrieval model to treat the target element characteristics of the case to be retrieved and each historical case. Historical element features are compared to determine the initial push case sequence from multiple historical cases; the push module is used to determine the target push case sequence of the case to be retrieved according to the initial push case sequence, so as to push the target push case sequence Push display of historical cases.

According to yet another aspect of the present disclosure, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the above-mentioned case retrieval and push method is implemented.

According to still another aspect of the present disclosure, there is provided an electronic device, including: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to execute the executable instructions to Execute the above case retrieval push method.

The case retrieval and push method provided by the embodiments of the present disclosure can first determine the target element characteristics of the cases to be retrieved and obtain the historical element characteristics of each historical case, then update the retrieval model based on the number of push times, and treat the case with the updated retrieval model The target element features of the retrieved cases are compared with the historical element features of each historical case, and the retrieval model is updated based on the number of push times, which can realize the implicit feedback of the results generated by the retrieval model, making the results of case retrieval push more accurate and more efficient. Can meet the user's personalized search needs.

It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure.

Description of drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the disclosure and together with the description serve to explain the principles of the disclosure. Apparently, the drawings in the following description are only some embodiments of the present disclosure, and those skilled in the art can obtain other drawings according to these drawings without creative efforts.

FIG. 1 shows a schematic diagram of an exemplary system architecture to which the case retrieval push method according to an embodiment of the present disclosure can be applied;

FIG. 2 shows a flow chart of a case retrieval push method according to an embodiment of the present disclosure;

FIG. 3 shows a flow chart of determining a target push case sequence in a case search and push method according to an embodiment of the present disclosure;

Fig. 4 shows a flow chart of determining an initial pushed case sequence from the plurality of historical cases through an updated retrieval model in the case retrieval push method according to an embodiment of the present disclosure;

FIG. 5 shows a flow chart of similarity calculation in a case retrieval push method according to an embodiment of the present disclosure;

FIG. 6 shows a schematic diagram of a case retrieval push method according to an embodiment of the present disclosure;

Fig. 7 shows a schematic diagram of implementing functional modules in a case retrieval push method according to an embodiment of the present disclosure;

FIG. 8 shows a block diagram of a case retrieval push device 700 according to an embodiment of the present disclosure; and

Fig. 9 shows a structural block diagram of a case retrieval push computer device in an embodiment of the present disclosure.

Detailed ways

Example embodiments will now be described more fully with reference to the accompanying drawings. Example embodiments may, however, be embodied in many forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this disclosure will be thorough and complete and will fully convey the concept of example embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

Furthermore, the drawings are merely schematic illustrations of the present disclosure and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and thus repeated descriptions thereof will be omitted. Some of the block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically separate entities. These functional entities may be implemented in software, or in one or more hardware modules or integrated circuits, or in different network and/or processor means and/or microcontroller means.

In addition, the terms "first" and "second" are used for descriptive purposes only, and cannot be interpreted as indicating or implying relative importance or implicitly specifying the quantity of indicated technical features. Thus, a feature defined as "first" and "second" may explicitly or implicitly include one or more of these features. In the description of the present disclosure, "plurality" means at least two, such as two, three, etc., unless otherwise specifically defined.

Fig. 1 shows a schematic diagram of an exemplary system architecture to which the case search and push method of the embodiment of the present disclosure can be applied.

As shown in FIG. 1 , the system architecture may include a server 101 , a network 102 and a client 103 . The network 102 is used as a medium for providing a communication link between the client 103 and the server 101 . Network 102 may include various connection types, such as wires, wireless communication links, or fiber optic cables, among others.

In an exemplary embodiment, the client 103 that performs data transmission with the server 101 may include, but not limited to, smart phones, desktop computers, tablet computers, notebook computers, smart speakers, digital assistants, and AR (Augmented Reality, Augmented Reality) devices , VR (Virtual Reality, virtual reality) equipment, smart wearable equipment and other types of electronic equipment. Optionally, the operating system running on the electronic device may include but not limited to Android system, IOS system, linux system, windows system and so on.

The server 101 can be an independent physical server, or a server cluster or a distributed system composed of multiple physical servers, and can also provide cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, Cloud servers for basic cloud computing services such as middleware services, domain name services, security services, CDN (Content Delivery Network, content distribution network), and big data and artificial intelligence platforms. In some practical applications, the server 101 may also be a server of a network platform. The network platform may be, for example, a transaction platform, a live broadcast platform, a social platform, or a music platform, which is not limited in this embodiment of the present disclosure. Wherein, the server may be one server, or may be a cluster formed by multiple servers, and the present disclosure does not limit the specific architecture of the server.

In an exemplary embodiment, the process by which the server 101 implements the push method for case retrieval may be as follows: the server 101 acquires the text data to be retrieved of the case to be retrieved; the server 101 processes the text data to be retrieved through the case element recognition model to obtain the case to be retrieved The target element features of each element; the server 101 obtains the historical element features of each element of multiple historical cases, and obtains the current number of pushes; the server 101 uses the search model based on the number of pushes, and the target element features of the case to be retrieved and each The historical element features of the historical cases are compared to determine the initial push case sequence from multiple historical cases; the server 101 determines the target push case sequence of the case to be retrieved according to the initial push case sequence, so as to push the target push case sequence Push display of historical cases.

In addition, it should be noted that what is shown in FIG. 1 is only an application environment of the case retrieval push method provided by the present disclosure. The numbers of clients, networks, and servers in FIG. 1 are only illustrative, and there may be any number of clients, networks, and servers according to actual needs.

In order to enable ordinary persons in the art to better understand the technical solution of the present disclosure, the steps of the case retrieval and push method in the exemplary embodiment of the present disclosure will be described in more detail below with reference to the drawings and embodiments.

Fig. 2 shows a flow chart of a case retrieval push method according to an embodiment of the present disclosure. The method provided by the embodiment of the present disclosure may be executed by the server or the client as shown in FIG. 1 , but the present disclosure is not limited thereto.

In the following illustrations, the server 101 is used as an execution subject for illustration.

As shown in Figure 2, the case retrieval push method provided by the embodiment of the present disclosure may include the following steps:

Step S201, acquiring the text data to be retrieved of the case to be retrieved.

The case to be retrieved may be an event involving litigation, etc., and the text data to be retrieved may be a text file or a judgment document used to describe the case to be retrieved.

In step S203, the text data to be retrieved is processed through the case element recognition model to obtain the target element features of each element of the case to be retrieved.

In some embodiments, the element features include at least one of the following: legal relationship features, core fact features, and evidentiary circumstances features. Among them, the case element recognition model can be pre-trained, which can be used to extract preset element information from the text, and convert the element information into element features; in some practical applications, element features can be written information.

For example, information extraction methods such as rule matching and deep learning can be used in the case element recognition model to identify and extract element features; Identify the keywords in the document, then extract the information within a certain range around the keywords, and then determine the characteristics of the elements. For deep learning, some sequence labeling algorithms or models can be used to extract feature features.

Step S205, acquiring the historical element features of each element of multiple historical cases, and acquiring the current number of pushed times.

In some practical applications, a case can be retrieved and calculated multiple times to be pushed, and the number of times pushed can be regarded as the number of times it has been displayed to the user for viewing.

In some embodiments, before obtaining the historical element features of each element of multiple historical cases, the case retrieval push method may further include: obtaining historical text data of multiple historical cases; processing each historical text data through a case element recognition model, The historical element characteristics of each historical case are obtained; the historical text data of the historical case and its historical element characteristics are correspondingly stored in the case database. Wherein, the historical text data may be information such as judgment documents of historical cases.

Based on this, step S205 may further include: obtaining historical element features of multiple historical cases from the above-mentioned constructed case database.

Step S207: Update the retrieval model based on the number of pushed times, compare the target element features of the case to be retrieved with the historical element features of each historical case through the updated retrieval model, and determine the initial push case sequence from multiple historical cases .

In this step, the similarity between the case to be retrieved and each historical case can be calculated according to the target element characteristics of the case to be retrieved and the historical element characteristics of each historical case, combined with the number of pushed times, and then the initial push case sequence can be determined. Among them, the user behavior feedback data can be determined according to the number of pushes, and then the retrieval model can be updated based on the user behavior feedback data, so as to realize the implicit feedback of user behavior on the results generated by the retrieval model, that is, the output can be dynamically adjusted according to the user's implicit feedback The search results make the results of the case search and push more accurate, and can better meet the personalized search needs of users.

Step S209, determine the target push case sequence of the cases to be retrieved according to the initial push case sequence, so as to push and display the historical cases in the target push case sequence.

In some embodiments, determining the target pushed case sequence of the case to be retrieved according to the initial pushed case sequence includes: if the number of times pushed is equal to 0, then using the initial pushed case sequence as the target pushed case sequence; if the number of times pushed is greater than 0, then Obtain the latest pushed history push case sequence, and determine the target push case sequence according to the historical push case sequence and the initial push case sequence.

Among them, the number of pushes equal to 0 can be understood as no push has been made before, and it is the first time to push the search results for the case to be retrieved, so the initial push case sequence obtained by calculation can be directly pushed as the target push case sequence . If the number of pushed cases is greater than 0, it can be understood that it has been pushed and displayed before, and the final target pushed case sequence for display can be determined by combining the previous historical pushed case sequence and the calculated initial pushed case sequence this time.

Among them, determining the target push case sequence according to the historical push case sequence and the initial push case sequence may be to determine the number and relative order of cases to be displayed in each sequence according to the two sequences, and then perform cross-recombination to obtain the target push case sequence.

In some practical applications, the historical cases clicked by the user in the last push can also be obtained. If there are historical cases clicked by the user in the initial push case sequence, the historical cases clicked by the user can be deleted first, and then the target can be determined. Push the case sequence, which can ensure that the cases in each push are not clicked and browsed by the user, which can improve the user's browsing and retrieval efficiency.

In some other practical applications, it is also possible to set the function of deleting the specified case or retaining the function of displaying the specified case, so as to meet the user's personalized search needs.

Fig. 3 shows a flow chart of determining the target push case sequence in the case retrieval and push method according to an embodiment of the present disclosure. Fig. 3 shows the case where the number of pushes has been greater than 0, as shown in Fig. 3 , including the historical push case sequence 301. The initial push case sequence 302 and the target push case sequence 303; where it is assumed that the number of cases in the history push case sequence 301 is 4, and the number of cases in the initial push case sequence 302 is 4; then the history push case sequence 301 can be pushed first Determine the subsequence composed of the top two cases in , then determine the subsequence composed of the top two cases from the initial push case sequence 302 , and then cross the cases in the two subsequences according to their respective relative orders Recombine to obtain a new sequence, that is, the target push case sequence 303 .

Through the case retrieval push method provided in this disclosure, the target element characteristics of the case to be retrieved and the historical element characteristics of each historical case can be obtained first, and then the retrieval model is updated based on the number of pushes, and the updated retrieval model is used to retrieve the case. The characteristics of the target elements are compared with the characteristics of the historical elements of each historical case, and the retrieval model is updated based on the number of times it has been pushed, which can realize the implicit feedback of the results generated by the retrieval model, making the results of case retrieval and push more accurate and more satisfying for users personalized search needs.

Fig. 4 shows a flow chart of determining the initial pushing case sequence from the multiple historical cases through the updated retrieval model in the case retrieval pushing method according to an embodiment of the present disclosure. As shown in Fig. 4 , in some embodiments , step S207 may further include the following steps.

Step S401, updating the current weight value of each element in the retrieval model based on the number of pushed times.

For example, in some embodiments, if the number of push times is equal to 0, the current weight value corresponding to each element is determined as the preset weight value; if the number of push times is greater than 0, the user behavior in the latest push is obtained Feedback data, and obtain the historical weight value of each element used in the retrieval model in the latest push; update the historical weight value according to the user behavior feedback data to obtain the current weight value of each element.

Among them, the number of push times equal to 0 can be understood as no push has been made before, and it is the first time to calculate the cases to be pushed for the cases to be retrieved, so it can be considered that the importance of each element is equivalent this time, and the corresponding elements can be The current weight values are all determined as preset weight values, for example, they are all set to 1. In the case that the number of push times is greater than 0, the user behavior feedback data in the latest push can be obtained to update the retrieval model.

In the case that the number of pushes has been greater than 0, in some embodiments, the process of obtaining the user behavior feedback data in the latest push may be: obtaining the user’s browsing time for each target historical case in the target push case sequence of the latest push, And obtain a preset duration threshold; convert the browsing duration into a behavior feedback value according to the duration threshold; determine the user behavior feedback data of each target historical case according to the behavior feedback value of each target historical case.

In some practical applications, the user behavior feedback data R can be in the following form:

R={(Hi,Ti),...}, i∈[1,...,N]; where, N is the number of output case results; Hi represents the click status of the i-th case, usually " Set "click" to 1, set "not click" to 0; Ti indicates the duration of the user's browsing in the i-th case, for example, it can be set to 4 states: 0 is not browsed, 1 is short-term browsed (such as the browsing time is usually Set at (0,3] minutes), 2 is medium-time browsing (such as browsing time is usually set at (3,10] minutes), 3 is long-time browsing (such as browsing time is usually set to be greater than 10 minutes); among them , the state setting of browsing time can be adjusted according to the actual situation. Among them, Ti can be regarded as the behavior feedback value of the i-th historical case, and (Hi,Ti) can be regarded as the user behavior feedback of the i-th historical case data.

In some embodiments, the process of updating the historical weight value according to the user behavior feedback data to obtain the current weight value of each element may be: obtaining the relationship between each target historical case and the case to be retrieved in the target push case sequence pushed last time The similarity between them is used to form the environmental data, and the historical weight is used as the action data; the environmental data, action data and user behavior feedback data are merged to obtain the input data; the input data is input into the trained deep reinforcement learning model , the output gets the current weight value of each element.

In some practical applications, the environment data s can be set as the similarity distribution s=[(s _i1 ,s _i2 ,...,s _iE ),...],i ∈[1,...,N], the action a can be the weight distribution a=(w ₁ ,w ₂ ,...,w _E ) used in the similarity calculation formula of the current retrieval model, where N is the output class The number of case results, and E is the number of elements set in the current business scenario.

Then the above (s, a, R) can be merged and input to the deep reinforcement learning model that has been trained offline, and the output will be the updated weight, that is, a′=(w ₁ ′,w ₂ ′,. .., w _E ').

Step S403, according to the current weight value of the target element features of the case to be retrieved and the historical element features of each historical case, the similarity calculation is performed to obtain the similarity between the case to be retrieved and each historical case.

Fig. 5 shows a flow chart of similarity calculation in the case retrieval push method according to an embodiment of the present disclosure. As shown in FIG. 5 , in some embodiments, the multiple historical cases may include the first case; the elements may include the first element; based on this, step S403 may include the following steps.

Step S501, using the similarity algorithm to calculate the similarity between the target element features of the first element of the case to be retrieved and the historical element features of the first element of the first case, and obtain the relationship between the case to be retrieved and the first element about the first element The element similarity of each element; thus the element similarity of each element between the case to be retrieved and the first case is obtained. Wherein, the similarity algorithm includes at least one of the following: Euclidean distance algorithm and cosine algorithm.

Step S503, perform weighted calculation according to the current weight value of each element and the corresponding element similarity, and obtain the similarity between the case to be retrieved and the first case.

Then, using the same method, the similarity between the case to be retrieved and each historical case can be obtained.

Step S405, sort the multiple historical cases according to the similarity, and obtain the sorting results of the multiple historical cases.

Step S407, determining the first preset number of historical cases in the sorting results as the initial pushing case sequence. Wherein, the value of the preset quantity can be set according to the actual situation or user's requirement.

It should be noted that the above-mentioned figures are only schematic illustrations of the processing included in the method according to the exemplary embodiment of the present invention, and are not intended to be limiting. It is easy to understand that the processes shown in the above figures do not imply or limit the chronological order of these processes. In addition, it is also easy to understand that these processes may be executed synchronously or asynchronously in multiple modules, for example.

Fig. 6 shows a schematic diagram of a case retrieval and pushing method according to an embodiment of the present disclosure, as shown in Fig. 6 , including the following steps:

Step 1, input the historical case document 601 into the case element recognition model 602, and use the case element recognition model 602 to extract the case element 603 of the historical case document 601 from the historical case document 601;

Step 2, storing the case elements 603 and corresponding documents (that is, historical case documents 601) generated in step 1 into the search database 606;

Step 3, input the retrieval case 604 to be retrieved into the case element recognition model 602, and use the case element recognition model 602 to extract the case element 605 of the retrieval case 604 from the retrieval case 604; wherein, in step 1 and step 3, use the case element The method for the identification model 602 to extract case elements can be: define E elements according to the business scenario of similar case retrieval, and the elements can include but not limited to core facts, legal relations, controversial legal issues, etc. and other information extraction methods for identification and extraction;

Step 4, input the case element 605 in step 3 and the data in the database 606 in step 2 into the retrieval model 607, and calculate the similarity between the retrieval case 604 and each historical case document 601, so that the historical case document 601 Retrieval and recall sorting; wherein, when the similarity calculation is performed for the first time, the weights corresponding to each element in the retrieval model 607 are the same;

其中s_e为检索案例和被检索案例关于第e个要素的相似度计算得分，w_e为第e个要素的权重，s_e的具体计算方式包括但不限于计算两个字符串之间的欧氏距离、基于词向量计算cosine距离等；Specifically, the retrieval model 607 involved in step 4 can calculate the similarity score between the retrieval case 604 and all historical case documents 601, and output the cases with the highest scores to the user; where, the initial calculation method of the similarity score is shown in formula:

Among them, s _e is the similarity calculation score between the retrieval case and the retrieved case on the e-th element, w _e is the weight of the e-th element, and the specific calculation method of s _e includes but is not limited to calculating the Euclidean value between two strings. Cosine distance, calculation of cosine distance based on word vector, etc.;

Step 5, when the results are displayed for the first time, output the top N similar case results generated in step 4 to the user;

Step 6, the user browses and clicks on the result in step 5 609, thereby generating an implicit feedback;

Step 7, the implicit feedback is passed to the agent module 610, and then the retrieval model 607 is updated by using the deep reinforcement learning algorithm DQN; wherein, specifically, the weight of each element in the retrieval model 607 can be updated;

Step 8, using the updated retrieval model 607 to recalculate the retrieval results, that is, to reorder the historical cases;

In step 9, the new results in step 8 can be cross-combined with the results generated in step 4, and the cases that the user has clicked before are filtered out and output to the user again;

Step 10, repeat steps 6, 7, 8, and 9 until the user finishes browsing the retrieval results.

Fig. 7 shows a schematic diagram of the implementation of functional modules in the case retrieval push method according to an embodiment of the present disclosure. As shown in Fig. 7, it includes: element extraction module 701, retrieval module 702, result output module 703 and reinforcement learning module 704;

Among them, the element extraction module 701 can extract several basic identification elements needed for similar cases from the case documents; the retrieval module 702 can retrieve the case elements to be retrieved from the case elements of the historical case documents, and retrieve the previous results with high similarity Output: The result output module 703 can display the results output from the retrieval module to the user, and record implicit feedback such as the user's click status and browsing time; the reinforcement learning module 704 can use the user's implicit feedback to dynamically adjust the retrieval module.

Fig. 8 shows a block diagram of a case retrieval push device 800 according to an embodiment of the present disclosure; as shown in Fig. 8 , it includes:

The acquisition module 801 is used to acquire the text data to be retrieved of the case to be retrieved; the processing module 802 is used to process the text data to be retrieved through the case element recognition model to obtain the target element features of each element of the case to be retrieved; the acquisition module 801 also uses To obtain the historical element features of each element of a plurality of historical cases, and obtain the current number of pushes; the processing module 802 is also used to update the retrieval model based on the number of pushes, and use the updated retrieval model to obtain the target element features and The historical element features of each historical case are compared to determine the initial push case sequence from multiple historical cases; the push module 803 is used to determine the target push case sequence of the case to be retrieved according to the initial push case sequence, so as to push the target Push the historical cases in the case sequence for push display.

Through the case retrieval and push device provided in this disclosure, it is possible to first determine the target element characteristics of the case to be retrieved and obtain the historical element characteristics of each historical case, and then update the retrieval model based on the number of push times, and use the updated retrieval model to retrieve the case. The characteristics of the target elements are compared with the characteristics of the historical elements of each historical case, and the retrieval model is updated based on the number of times it has been pushed, which can realize the implicit feedback of the results generated by the retrieval model, making the results of case retrieval and push more accurate and more satisfying for users personalized search needs.

In some embodiments, the processing module 802 updates the retrieval model based on the number of times it has been pushed, and compares the target element features of the case to be retrieved with the historical element features of each historical case through the updated retrieval model, so as to extract from multiple historical cases Determine the initial push case sequence, including: update the current weight value of each element in the retrieval model based on the number of pushes; calculate the similarity between the target element features and each historical element feature according to the current weight value, and obtain the case to be retrieved and each historical case. The similarity between multiple historical cases is sorted according to the similarity to obtain the sorting results of multiple historical cases; the previous preset number of historical cases in the sorting results are determined as the initial push case sequence.

In some embodiments, the processing module 802 updates the current weight value of each element in the retrieval model based on the number of push times, including: if the number of push times is equal to 0, then determining the current weight value corresponding to each element as a preset weight value; If the number of pushes has been greater than 0, the acquisition module 801 obtains the user behavior feedback data in the latest push, and obtains the historical weight value of each element used in the retrieval model in the latest push; The weight values are updated to get the current weight values for each feature.

In some embodiments, the obtaining module 801 obtains the user behavior feedback data in the latest push, including: obtaining the user's browsing time for each target historical case in the target push case sequence of the latest push, and obtaining a preset duration threshold; The browsing time is converted into a behavior feedback value according to the duration threshold; the user behavior feedback data of each target historical case is determined according to the behavior feedback value of each target historical case.

In some embodiments, the processing module 802 updates the historical weight value according to the user behavior feedback data to obtain the current weight value of each element, including: the obtaining module 801 obtains the target historical case and The similarity between the cases to be retrieved is used to form the environmental data, and the historical weight is used as the action data; the processing module 802 merges the environmental data, action data and user behavior feedback data to obtain the input data; input the input data to the training In a good deep reinforcement learning model, the output gets the current weight value of each element.

In some embodiments, the multiple historical cases include the first case; the elements include the first element; wherein, the processing module 802 calculates the similarity between the target element feature and each historical element feature according to the current weight value, and obtains the case to be retrieved and each The similarity between historical cases includes: using a similarity algorithm to calculate the similarity between the target element feature of the first element of the case to be retrieved and the historical element feature of the first element of the first case, and obtain the case to be retrieved and the first element The element similarity of the first element between the cases, so as to obtain the element similarity of each element between the case to be retrieved and the first case; carry out weighted calculation according to the current weight value of each element and the corresponding element similarity, and obtain the to-be-retrieved case The similarity between the retrieved case and the first case is retrieved.

In some embodiments, the push module 803 determines the target push case sequence of the case to be retrieved according to the initial push case sequence, including: if the push count is equal to 0, then use the initial push case sequence as the target push case sequence; if the push count is greater than 0, the latest push history push case sequence is obtained, and the target push case sequence is determined according to the history push case sequence and the initial push case sequence.

In some embodiments, the acquisition module 801 is also used to: acquire historical text data of multiple historical cases; the processing module 802 is also used to: process each historical text data through a case element recognition model to obtain the historical element features of each historical case; The historical text data of historical cases and their historical element features are correspondingly stored in the case database; and the acquisition module 801 acquires the historical element features of multiple historical cases, including: acquiring the historical element features of multiple historical cases from the case database.

In some embodiments, the element features include at least one of the following: legal relationship features, core fact features, and evidentiary circumstances features.

Other content of the embodiment in FIG. 8 may refer to the above-mentioned other embodiments.

Those skilled in the art can understand that various aspects of the present invention can be implemented as systems, methods or program products. Therefore, various aspects of the present invention can be embodied in the following forms, that is: a complete hardware implementation, a complete software implementation (including firmware, microcode, etc.), or a combination of hardware and software implementations, which can be collectively referred to herein as "circuit", "module" or "system".

Fig. 9 shows a structural block diagram of a case retrieval push computer device in an embodiment of the present disclosure. It should be noted that the electronic device shown in the figure is only an example, and should not impose any limitation on the functions and application scope of the embodiments of the present invention.

An electronic device 900 according to this embodiment of the present invention is described below with reference to FIG. 9 . The electronic device 900 shown in FIG. 9 is only an example, and should not limit the functions and scope of use of this embodiment of the present invention.

As shown in FIG. 9, electronic device 900 takes the form of a general-purpose computing device. Components of the electronic device 900 may include but not limited to: at least one processing unit 910 , at least one storage unit 920 , and a bus 930 connecting different system components (including the storage unit 920 and the processing unit 910 ).

Wherein, the storage unit stores program codes, and the program codes can be executed by the processing unit 910, so that the processing unit 910 executes various exemplary methods according to the present invention described in the "Exemplary Methods" section of this specification. Implementation steps. For example, the processing unit 910 may execute the method as shown in FIG. 2 .

The storage unit 920 may include a readable medium in the form of a volatile storage unit, such as a random access storage unit (RAM) 9201 and/or a cache storage unit 9202 , and may further include a read-only storage unit (ROM) 9203 .

The storage unit 920 may also include a program/utility 9204 having a set (at least one) of program modules 9205, such program modules 9205 including but not limited to: an operating system, one or more application programs, other program modules, and program data, Implementations of networked environments may be included in each or some combination of these examples.

Bus 930 may represent one or more of several types of bus structures, including a memory unit bus or memory unit controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local area using any of a variety of bus structures. bus.

The electronic device 900 can also communicate with one or more external devices 1000 (such as keyboards, pointing devices, Bluetooth devices, etc.), and can also communicate with one or more devices that enable the user to interact with the electronic device 900, and/or communicate with Any device (eg, router, modem, etc.) that enables the electronic device 900 to communicate with one or more other computing devices. Such communication may occur through input/output (I/O) interface 950 . Moreover, the electronic device 900 can also communicate with one or more networks (such as a local area network (LAN), a wide area network (WAN) and/or a public network such as the Internet) through the network adapter 960 . As shown, the network adapter 960 communicates with other modules of the electronic device 900 through the bus 930 . It should be appreciated that although not shown, other hardware and/or software modules may be used in conjunction with electronic device 900, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives And data backup storage system, etc.

In an exemplary embodiment of the present disclosure, there is also provided a computer-readable storage medium on which a program product capable of implementing the above-mentioned method in this specification is stored. In some possible implementations, various aspects of the present invention can also be implemented in the form of a program product, which includes program code, and when the program product is run on a terminal device, the program code is used to make the The terminal device executes the steps according to various exemplary embodiments of the present invention described in the "Exemplary Method" section above in this specification.

According to the program product for implementing the above method according to the embodiment of the present invention, it may adopt a portable compact disk read-only memory (CD-ROM) and include program codes, and may run on a terminal device such as a personal computer. However, the program product of the present invention is not limited thereto. In this document, a readable storage medium may be any tangible medium containing or storing a program, and the program may be used by or in combination with an instruction execution system, apparatus or device.

The program product may reside on any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or device, or any combination thereof. More specific examples (non-exhaustive list) of readable storage media include: electrical connection with one or more conductors, portable disk, hard disk, random access memory (RAM), read only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

A computer readable signal medium may include a data signal carrying readable program code in baseband or as part of a carrier wave. Such propagated data signals may take many forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the foregoing. A readable signal medium may also be any readable medium other than a readable storage medium that can transmit, propagate, or transport a program for use by or in conjunction with an instruction execution system, apparatus, or device.

Program code embodied on a readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

Program code for carrying out the operations of the present invention may be written in any combination of one or more programming languages, including object-oriented programming languages—such as Java, C++, etc., as well as conventional procedural programming languages. Programming language - such as "C" or a similar programming language. The program code may execute entirely on the user's computing device, partly on the user's device, as a stand-alone software package, partly on the user's computing device and partly on a remote computing device, or entirely on the remote computing device or server to execute. In cases involving a remote computing device, the remote computing device may be connected to the user computing device through any kind of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., using an Internet service provider). business to connect via the Internet).

It should be noted that although several modules or units of the device for action execution are mentioned in the above detailed description, this division is not mandatory. Actually, according to the embodiment of the present disclosure, the features and functions of two or more modules or units described above may be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided to be embodied by a plurality of modules or units.

In addition, although steps of the methods of the present disclosure are depicted in the drawings in a particular order, there is no requirement or implication that the steps must be performed in that particular order, or that all illustrated steps must be performed to achieve the desired result. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into one step for execution, and/or one step may be decomposed into multiple steps for execution, etc.

Through the description of the above implementations, those skilled in the art can easily understand that the example implementations described here can be implemented by software, or by combining software with necessary hardware. Therefore, the technical solutions according to the embodiments of the present disclosure can be embodied in the form of software products, and the software products can be stored in a non-volatile storage medium (which can be CD-ROM, U disk, mobile hard disk, etc.) or on the network , including several instructions to make a computing device (which may be a personal computer, a server, a mobile terminal, or a network device, etc.) execute the method according to the embodiments of the present disclosure.

According to an aspect of the present disclosure there is provided a computer program product or computer program comprising computer instructions stored in a computer readable storage medium. The processor of the computer device reads the computer instruction from the computer-readable storage medium, and the processor executes the computer instruction, so that the computer device executes the methods provided in various optional implementation manners of the foregoing embodiments.

Other embodiments of the present disclosure will be readily apparent to those skilled in the art from consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any modification, use or adaptation of the present disclosure, and these modifications, uses or adaptations follow the general principles of the present disclosure and include common knowledge or conventional technical means in the technical field not disclosed in the present disclosure . The specification and examples are to be considered exemplary only, with the true scope and spirit of the disclosure indicated by the appended claims.

Claims (12)

1. A case retrieval pushing method is characterized by comprising the following steps:

acquiring text data to be retrieved of a case to be retrieved;

processing the text data to be retrieved through a case element identification model to obtain target element characteristics of each element of the case to be retrieved;

acquiring historical element characteristics of each element of a plurality of historical cases, and acquiring the current pushed times;

updating a retrieval model based on the pushed times, and comparing the target element characteristics of the case to be retrieved with the historical element characteristics of each historical case through the updated retrieval model to determine an initial pushed case sequence from the plurality of historical cases;

and determining a target pushing case sequence of the case to be retrieved according to the initial pushing case sequence so as to push and display the historical cases in the target pushing case sequence.

2. The method as claimed in claim 1, wherein updating a retrieval model based on the pushed times, and comparing the target element features of the case to be retrieved with the historical element features of each historical case through the updated retrieval model to determine an initial pushed case sequence from the plurality of historical cases comprises:

updating the current weight value of each element in the retrieval model based on the pushed times;

according to the current weight value, similarity calculation is carried out on the target element characteristics and the historical element characteristics to obtain the similarity between the case to be retrieved and each historical case;

sequencing the plurality of historical cases according to the similarity to obtain sequencing results of the plurality of historical cases;

and determining the previous preset number of historical cases in the sequencing result as the initial push case sequence.

3. The method of claim 2, wherein updating the current weight values of the elements in the search model based on the pushed times comprises:

if the pushed times are equal to 0, determining the current weight values corresponding to the elements as preset weight values;

if the pushed times are larger than 0, user behavior feedback data in the latest pushing is obtained, and historical weight values of all elements used in the retrieval model in the latest pushing are obtained; and updating the historical weight value according to the user behavior feedback data to obtain the current weight value of each element.

4. The method of claim 3, wherein obtaining the user behavior feedback data in the last push comprises:

acquiring the browsing duration of each target historical case in a target pushing case sequence pushed by a user for the last time, and acquiring a preset duration threshold;

converting the browsing duration into a behavior feedback value according to the duration threshold;

and determining the user behavior feedback data of each target history case according to the behavior feedback value of each target history case.

5. The method of claim 3, wherein updating the historical weight values according to the user behavior feedback data to obtain current weight values of the elements comprises:

acquiring similarity between each target historical case in a target pushing case sequence which is pushed for the last time and the case to be retrieved to form environment data, and taking the historical weight as action data;

merging the environment data, the action data and the user behavior feedback data to obtain input data;

and inputting the input data into a trained deep reinforcement learning model, and outputting to obtain the current weight value of each element.

6. The method of claim 2, wherein said plurality of historical cases includes a first case; the element comprises a first element; wherein, according to the current weight value, performing similarity calculation on the target element feature and each historical element feature to obtain the similarity between the case to be retrieved and each historical case, and the method comprises the following steps:

performing similarity calculation on the target element characteristics of the first element of the case to be retrieved and the historical element characteristics of the first element of the first case by adopting a similarity calculation method to obtain the element similarity of the first element between the case to be retrieved and the first case, so as to obtain the element similarity of each element between the case to be retrieved and the first case;

and performing weighting calculation according to the current weight value of each element and the corresponding element similarity to obtain the similarity between the case to be retrieved and the first case.

7. The method of claim 1, wherein determining a target push case sequence of the case to be retrieved from the initial push case sequence comprises:

if the pushed times are equal to 0, taking the initial pushed case sequence as the target pushed case sequence;

and if the pushed times are larger than 0, acquiring a historical pushed case sequence pushed for the last time, and determining the target pushed case sequence according to the historical pushed case sequence and the initial pushed case sequence.

8. The method of any of claims 1-7, further comprising:

acquiring historical text data of the plurality of historical cases;

processing each historical text data through the case element recognition model to obtain the historical element characteristics of each historical case;

correspondingly storing the historical text data of the historical case and the historical element characteristics thereof to a case database;

in the method, acquiring the historical element characteristics of a plurality of historical cases comprises the following steps: and acquiring historical element characteristics of the plurality of historical cases from the case database.

9. The method according to any one of claims 1-7, wherein the feature characteristics include at least one of: legal relations characteristics, core facts characteristics, and proof cases characteristics.

10. A case retrieval pushing device is characterized by comprising:

the acquisition module is used for acquiring text data to be retrieved of the case to be retrieved;

the processing module is used for processing the text data to be retrieved through a case element identification model to obtain target element characteristics of each element of the case to be retrieved;

the acquisition module is also used for acquiring the historical element characteristics of each element of a plurality of historical cases and acquiring the current pushed times;

the processing module is also used for updating a retrieval model based on the pushed times, and comparing the target element characteristics of the case to be retrieved with the historical element characteristics of each historical case through the updated retrieval model so as to determine an initial pushed case sequence from the plurality of historical cases;

and the pushing module is used for determining a target pushing case sequence of the case to be retrieved according to the initial pushing case sequence so as to push and display the historical cases in the target pushing case sequence.

11. A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out a case retrieval pushing method according to any one of claims 1 to 9.

12. An electronic device, comprising:

one or more processors;

storage means for storing one or more programs which, when executed by the one or more processors, cause the one or more processors to implement the case retrieval pushing method of any one of claims 1 to 9.

Images