JP7540290B2 - Dialogue robot system, dialogue method and dialogue program - Google Patents

Description

The present invention relates to a dialogue robot system equipped with a dialogue robot terminal, a dialogue method used in the dialogue robot system, and a dialogue program for causing a computer to execute the dialogue method.

Conventionally, conversational robots that output responses to user utterances by voice are known. For example, the conversational robot disclosed in Patent Document 1 is connected to a server. When a user speaks to the conversational robot, the server analyzes the meaning of the speech, refers to a response sentence database to select a response sentence that corresponds to the meaning, and transmits the response sentence in the form of voice data to the conversational robot. The conversational robot outputs the response sentence to the user by voice.

Patent No. 6633250

It is believed that such conversational robots can also be suitably applied to providing guidance services that answer users' questions. For example, if conversational robots were installed in place of station staff at multiple railway stations operated by a single railway operator, it would be possible to both reduce the labor required by station staff and maintain or improve guidance services. However, if the conventional conversational robots described above were installed as is at each railway station, the following problems would likely arise.

Specific examples of guidance at railway stations include information on how to purchase a train ticket, and station facilities such as restroom locations and store opening hours. With regard to guidance on how to purchase a train ticket, it may be possible to standardize the answers to the same questions across all railway stations. On the other hand, with regard to station facilities, it is difficult to standardize the answers to the same questions.
Therefore, it may be necessary to install a server at each of the multiple railway stations where the conversational robot is installed, and have each server store a database that defines the correspondence between questions and answers specific to that railway station.

This server is a so-called "dialogue engine" that processes information by analyzing the meaning of utterances and selecting response sentences. The work of building and maintaining the dialogue engine must be done for each railway station. If the number of stations where dialogue robots are installed increases to save on the labor of station staff, this will in turn lead to an increase in the workload.
It is believed that such problems will occur in the same way when multiple conversational robots are installed in multiple locations and it is difficult to standardize responses to the same questions.

The present invention aims to reduce the workload involved in building and maintaining a dialogue robot system equipped with multiple dialogue robot terminals.

A dialogue robot system according to one embodiment of the present invention includes a plurality of dialogue robot terminals that output answer information in response to a user's utterance, a centralized processing device communicatively connected to the plurality of dialogue robot terminals, a dialogue database provided in the centralized processing device, and a plurality of answer databases provided in each of the plurality of dialogue robot terminals. The dialogue database defines the correspondence between question contents and answer identifiers included in the utterance, and the answer database defines the correspondence between answer identifiers and answer information. The dialogue robot terminal outputs utterance information indicating the contents of the utterance to the centralized processing device. The centralized processing device identifies the question contents from the utterance information output from the dialogue robot terminal, determines an answer identifier corresponding to the identified question contents according to the dialogue database, and outputs the determined answer identifier to the dialogue robot terminal. The dialogue robot terminal determines the answer information corresponding to the output answer identifier according to the answer database.

Here, if a dialogue robot system has multiple dialogue robot terminals, it may not be possible to standardize the answer content for the same question across all dialogue robot terminals. If the dialogue engine were to complete the determination of the specific answer content as in the past, it would be necessary to prepare multiple dialogue engines for each of the multiple dialogue robot terminals. This would increase the workload of building and maintaining the dialogue engines.

According to the above configuration, multiple interactive robot terminals are connected to a centralized processing device. The centralized processing device is provided with an interactive database that defines the correspondence between question contents and answer identifiers, and each interactive robot terminal is provided with an answer database that defines the correspondence between answer identifiers and answer information. When a user speaks to a interactive robot terminal, it transmits speech information indicating the content of the question to the centralized processing device. The centralized processing device identifies the question content from the speech information and returns an answer identifier corresponding to the question content to the interactive robot terminal. The interactive robot terminal determines the answer information to be output to the user based on the answer identifier.

The centralized processing device has the function of a "dialogue engine" that performs information processing by analyzing speech information, identifying the question, and selecting an answer, but does not go so far as to determine the specific content of the answer. The specific content of the answer output to the user is determined by each dialogue robot terminal. Two types of databases are provided separately for the centralized processing device and the dialogue robot terminal, and an answer identifier is mediated in the process of determining the answer content from the question content. It is sufficient that each of the multiple answer databases stores information specific to the dialogue robot terminal as answer information corresponding to the answer identifier.

This allows the dialogue engine to be shared among multiple dialogue robot terminals while outputting unique answer information to the user, thus reducing the workload of building and maintaining the dialogue engine even if the number of dialogue robot terminals constituting the system increases.
The plurality of dialogue robot terminals may each have a plurality of answer databases.
Each of the interactive robot terminals may input the contents of a user's utterance as voice data, and output text data converted from the input voice data to the centralized processing device as utterance information.

Generally, the amount of text data is smaller than that of voice data. Therefore, the amount of information to be transmitted from the interactive robot terminal to the centralized processing device is small. The centralized processing device can identify the content of the question based on the text data, and can realize rapid identification processing.
Each of the interactive robot terminals may be configured to determine answer information as text data based on the answer identifier, and may output voice data converted from the text data as the answer information.

According to the above configuration, even if the answer information stored in the answer database is text data, the text data is converted into voice data, so that the answer information can be spoken out loud to realize a dialogue with the user.
The answer database may store answer information corresponding to answer identifiers in a plurality of languages. The interactive robot terminal may output the answer information in the same language as the utterance.

According to the above configuration, it is possible to realize a dialogue with users who speak various languages.
A plurality of interactive robot terminals may be installed in each of a plurality of railway stations.
According to the above configuration, it is possible to have a dialogue with the user at the railway station, which makes it possible to both reduce the railway operator's guidance work and enrich the information provided to the user.
A dialogue method according to one embodiment of the present invention is used in a dialogue robot system including a plurality of dialogue robot terminals that output answer information in response to a user's utterance, and a centralized processing device communicably connected to the plurality of dialogue robot terminals. The dialogue method includes outputting utterance information indicating the content of the utterance from the dialogue robot terminal to the centralized processing device, identifying a question content from the output utterance information in the centralized processing device, determining an answer identifier corresponding to the identified question content in accordance with a dialogue database that defines a correspondence between the question content included in the utterance and an answer identifier in the centralized processing device, outputting the determined answer identifier from the centralized processing device to the dialogue robot terminal, and determining, in the dialogue robot terminal, the answer information corresponding to the output answer identifier in accordance with the answer database that defines a correspondence between the answer identifier and the answer information.

A dialogue program according to one embodiment of the present invention causes a computer to execute a dialogue method used in a dialogue robot system including a plurality of dialogue robot terminals that output answer information in response to a user's utterance, and a central processing device communicably connected to the plurality of dialogue robot terminals. The dialogue program includes outputting speech information indicating the content of the utterance from the dialogue robot terminal to the central processing device, identifying a question content from the output speech information in the central processing device, determining an answer identifier corresponding to the identified question content according to a dialogue database that defines the correspondence between the question content included in the utterance and an answer identifier in the central processing device, outputting the determined answer identifier from the central processing device to the dialogue robot terminal, and determining answer information corresponding to the output answer identifier in the dialogue robot terminal according to the answer database that defines the correspondence between the answer identifier and answer information.

The above dialogue method and dialogue program have technical features corresponding to the features of the dialogue robot system described above, and provide the same effects as the dialogue robot system.

The present invention can reduce the workload involved in building and maintaining a dialogue robot system equipped with multiple dialogue robot terminals.

1 is a configuration diagram of a dialogue robot system according to an embodiment of the present invention. FIG. 2 is a diagram showing a dialogue database and a response database. FIG. 2 is a block diagram showing an interactive robot terminal and a centralized processing device. 2 is a flowchart showing a dialogue method and a dialogue program according to an embodiment of the present invention; FIG. 1 is a configuration diagram of an interactive robot system according to a reference example.

Hereinafter, an embodiment of the present invention will be described with reference to the drawings. In addition, the same or corresponding elements are designated by the same reference numerals throughout the drawings, and detailed description thereof will be omitted.
1 is a configuration diagram of a dialogue robot system 1 according to an embodiment of the present invention. The dialogue robot system 1 includes a plurality of dialogue robot terminals 2 (2A, 2B, ...) and a central processing device 3. Each dialogue robot terminal 2 is connected to the central processing device 3 via a communication network 9 so as to be able to communicate with each other.

In this embodiment, the interactive robot system 1 is used for guidance services at station counters operated by a railway operator. In this case, multiple interactive robot terminals 2A, 2B, ... are installed at multiple railway stations owned by the railway operator. In the illustrated example, a first interactive robot terminal 2A is installed at station A, a second interactive robot terminal 2B is installed at station B, and a third interactive robot terminal 2C is installed at station C. However, the number of interactive robot terminals 2 that make up the interactive robot system 1 is not particularly limited.

Each interactive robot terminal 2 has a housing 20. Each interactive robot terminal 2 further has a controller 21, a microphone 26, a speaker 27, and a display 28, which are attached inside, on the surface, or outside of the housing 20. The interactive robot terminal 2 may further have a camera (not shown) that captures images of the surroundings, and an alarm (not shown) that sounds when an abnormality occurs.

The controller 21 has a CPU 22, a main memory 23, and an interface 24. The CPU 22 executes part of the dialogue method according to the present embodiment in accordance with a terminal program P2 prestored in the main memory 23. The main memory 23 is composed of storage devices such as ROM, RAM, and EEPROM, and can also temporarily store information required to execute the dialogue method. The interface 24 is connected to a microphone 26, a speaker 27, and a display 28. The interface 24 transmits and receives information to and from the centralized processing device 3 via the communication network 9, and also transmits and receives information to and from an external application 90 as necessary.

The microphone 26 collects the speech of the user U. The speaker 27 can output various audio information. The display 28 can display various image information and text information. As an example, the speaker 27 outputs response information to the speech of the user U in the form of audio. The display 28 outputs response information to the speech of the user U in the form of text display.

The centralized processing device 3 is realized by, for example, a server device. The centralized processing device 3 is a so-called “dialogue engine” that performs information processing to identify the meaning of an utterance from the user U to the interactive robot terminal 2 and prepare a response to the user U.
The centralized processing device 3 has a controller 31. The controller 31 has a CPU 32, a main memory 33, and an interface 34. The CPU 32 executes a part of the dialogue method according to the present embodiment in accordance with an engine program P3 prestored in the main memory 33. The interface 34 inputs and outputs information to and from the multiple dialogue robot terminals 2 constituting the system 1 via the communication network 9, and also inputs and outputs information to and from an external application 90 as necessary.

The CPU 22 and the CPU 32 (or the controller 21 and the controller 31 respectively having these) are an example of a computer that executes the dialogue method according to this embodiment. The terminal program P2 and the engine program P3 constitute a dialogue program P that causes a computer to execute this dialogue method.
In such an interactive robot system 1, each interactive robot terminal 2 outputs response information in response to an utterance by a user U. For example, assume that the user U speaks to one of the interactive robot terminals 2, saying, "Where's the toilet?" or "I want to go to the toilet." The user U's utterance is collected by the microphone 26. The interactive robot terminal 2 outputs speech information indicating the content of the utterance to the centralized processing device 3. The centralized processing device 3 identifies the content of the user U's question from the speech information as a question about "where the toilet is."

Because the structure or layout of station facilities differs from station to station, the relative position of the toilet to the interactive robot terminal 2 may also differ. Therefore, it is necessary to output station-specific answer information for the same question. The first interactive robot terminal 2A outputs answer information that is appropriate for Station A, such as "The toilet is on the left," while the second interactive robot terminal 2B outputs answer information that is appropriate for Station B, such as "The toilet is on the right." The answer information is output in the form of speech from at least the speaker 27, and is also output in the form of text display on the display 28. This allows a conversation to be established with the user U, and an appropriate guidance service is provided to the user U in an unmanned manner.

The centralized processing device 3, which serves as a dialogue engine, identifies the question content from the utterance information. The process of determining station-specific answer information from the identified question content is shared between the centralized processing device 3 and the dialogue robot terminal 2. The dialogue robot terminal 2 shares the downstream process of determining the answer information, and outputs the answer information determined by itself to the user U.
To realize this sharing of processing, the interactive robot system 1 includes an interactive database 35 provided in the centralized processing device 3, and a plurality of answer databases 25 (25A, 25B, ...) provided corresponding to each of the plurality of interactive robot terminals 2 (2A, 2B, ...). Each of the plurality of interactive robot terminals 2 has a plurality of answer databases 25. Although FIG. 1 shows that the answer database 25 is provided outside the housing 20 for convenience of illustration, the answer database 25 may be provided inside the housing 20.

Hereinafter, the database corresponding to the first interactive robot terminal 2A installed at station A is referred to as the first response database 25A. The second response database 25B and the third response database 25C correspond to the second interactive robot terminal 2B and the third interactive robot terminal 2C, respectively.
FIG. 2 shows an example of the dialogue database 35 and the answer database 25. The dialogue database 35 defines the correspondence between the question content included in the utterance and the answer identifier. In FIG. 2, as a mere example, seven question items correspond one-to-one with different answer identifiers in the dialogue database 35. This interactive robot system 1 can answer the user U for these seven question items preset in the dialogue database 35. Note that, as a measure in case the question content cannot be identified, the dialogue database 35 also records an answer identifier corresponding to "unidentifiable" (see "No. 8" on the last line).

Information about train tickets and station facilities is often provided to user U at train stations. The five questions, "No. 1 to 5," are about train tickets. In general, rules related to train tickets are uniformly applied to all train stations operated by a railway operator that has introduced the interactive robot system 1. Therefore, it may be possible to standardize the answers to the same questions at all train stations.

On the other hand, the two questions "No. 6-7" surrounded by a thick line are related to station facilities. Since railway stations vary in structure and size, and the locations of the interactive robot terminals 2 vary from station to station, the answers to the same questions cannot be standardized across all railway stations. In other words, the answers will be station-specific. Note that questions for which the answers may be station-specific include information about station facilities, as well as information about tourist facilities around the station and information about travel routes (transfers) to the destination.

In Fig. 2, the answer identifiers are represented by symbols such as "Q1", "Q2" ..., but this is merely an example. The answer identifiers may be any information or data that differs from one another. In this example, it is sufficient to prepare eight answer identifiers, one more than the number of questions, and each answer identifier may be simplified to a small piece of data of three bits (three binary digits).
The answer database 25 defines the correspondence between answer identifiers and answer information. The answer information defined in each answer database 25 corresponds to the question content defined in the dialogue database 35 via the answer identifier. In other words, the question content defined in the dialogue database 35 corresponds to multiple pieces of answer information defined in each of the multiple answer databases 25 via the answer identifier.

For example, answer identifier "Q1" corresponds to the question "About children's tickets" in the dialogue database 35. The question "About children's tickets" is information about train tickets, and the answer can be made common to all stations. Therefore, in each answer database 25 (25A, 25B, ...), answer identifier "Q1" corresponds to the same answer information "That's the ticket fare for children." The same is true for answer identifiers "Q2" to "Q5."

On the other hand, the answer identifier "Q6" corresponds to the question "Where is the toilet?" in the dialogue database 35. The question "Where is the toilet?" is information about station facilities, and the answer cannot be made common to all stations. In the first answer database 25A, the answer identifier "Q6" corresponds to the answer information "The toilet is on the left" in light of the structure of station A and the installation position of the first dialogue robot terminal 2A. In the second answer database 25B, the answer identifier "Q6" corresponds to the answer information "The toilet is on the right" in light of the structure of station B and the installation position of the second dialogue robot terminal 2B. Although detailed illustration is omitted, in the third answer database 25C, the answer identifier "Q6" corresponds to the answer information suitable for station C. The same is true for the answer identifier "Q7". In each answer database 25, station-specific answer information to the question "About the opening hours of the convenience store" is recorded.

Each answer database 25 has a column for answer identifier and answer information, as well as a column for language. In the figure, the information indicating the language is represented by a two-letter alphabetic code. "ja" indicates Japanese and "en" indicates English, but the languages handled by the interactive robot terminal 2 are not limited to these two.
A plurality of correspondence relationships between answer identifiers and answer information are defined for each language. If the number of question contents and answer identifiers is n and the number of languages handled by the interactive robot terminal 2 is m, n×m pieces of answer information are registered in each answer database 25.

On the other hand, in the dialogue database 35, n question contents and answer identifiers are registered regardless of the number of languages handled by the dialogue robot terminal 2. In other words, the centralized processing device 3 as a dialogue engine handles one specific language (for example, Japanese).
3 is a block diagram showing the interactive robot terminal 2 and the centralized processing device 3. The interactive robot terminal 2 executes the terminal program P2 (see FIG. 1) and thereby has a data adjustment unit 41, an inquiry unit 42, a reply reception unit 43, and a reply information determination unit 44. The centralized processing device 3 executes the engine program P3 and thereby has an inquiry reception unit 51, a question content specification unit 52, a reply identifier determination unit 53, and a reply unit 54.

Figure 4 is a flowchart showing the dialogue method according to this embodiment and the dialogue program P for causing a computer to execute the method. Below, the operation of the configuration shown in Figure 3 will be explained along with the steps of the dialogue method shown in Figure 4. The dialogue method starts when the user U speaks to the dialogue robot terminal 2, and ends when the dialogue robot terminal 2 outputs answer information to the user U.

When the user U speaks to the interactive robot terminal 2, the data adjustment unit 41 detects that speech has been made based on the voice collected by the microphone 26 (S11). Next, the data adjustment unit 41 identifies the language spoken by the user U based on the voice data acquired by the microphone 26 (S13), and acquires text data converted from the voice data (S14).
A known automatic language identification algorithm and data conversion algorithm using a neural network are applied for language identification and conversion from voice data to text data. The number of languages that can be identified by the automatic language identification algorithm is equal to the number of languages that can be handled by the dialogue robot terminal 2. The automatic language identification algorithm may be incorporated in the terminal program P2 or may be provided by an external application 90.

In other words, the data adjustment unit 41 may perform language identification processing and data conversion processing within the interactive robot terminal 2. Alternatively, the data adjustment unit 41 may identify the language by transmitting voice data to the external application 90 after detecting speech and receiving the identification result obtained by the external application 90 from the external application 90. At the same time, the data adjustment unit 41 may obtain the text data converted by the external application 90 from the external application 90.

The converted text data is written in the language of the speech. The data adjustment unit 41 determines whether the identified language is Japanese or not (S14). If it is not Japanese (S14: N), the data adjustment unit 41 acquires the text data translated into Japanese (S15). A known machine translation algorithm is applied to translate the text data. The machine translation algorithm may be built into the terminal program P2, or may be provided by an external application 90.

In other words, the data adjustment unit 41 may perform machine translation processing within the interactive robot terminal 2. Alternatively, the data adjustment unit 41 may transmit text data written in a language other than Japanese to the external application 90, and obtain from the external application 90 the text data translated into Japanese by the external application 90.
The processes (S11 to S15) performed in the interactive robot terminal 2 up to this point are pre-processing prior to the process of identifying answer information corresponding to the utterance information. After this pre-processing, the inquiry unit 42 outputs utterance information indicating the content of the utterance of the user U to the centralized processing device 3 (S16).

The inquiry unit 42 outputs the Japanese text data acquired by the preprocessing as the utterance information, and requests the centralized processing device 3 to analyze the utterance information. At this time, the process performed in the interactive robot terminal 2 is temporarily suspended, and the centralized processing device 3 continues to execute the interactive method.
In the centralized processing device 3, first, the inquiry receiving unit 51 receives the output speech information (S31). Then, the question content identifying unit 52 identifies the question content from the output speech information (Japanese text data) (S32). In other words, the question content identifying unit 52 identifies which of a plurality of question contents preset in the dialogue database 35 the output speech information includes. A known analysis algorithm using a neural network is applied to the process of analyzing the text data and identifying its meaning (question content). The analysis algorithm may be incorporated in the engine program P3 or may be provided by the external application 90.

In other words, the question content identification unit 52 may perform the analysis process within the centralized processing device 3. Alternatively, the question content identification unit 52 may transmit utterance information (Japanese text data) to the external application 90 and receive an analysis result of the external application 90 from the external application 90, thereby identifying the question content.
Next, the answer identifier determination unit 53 determines an answer identifier corresponding to the identified question content in accordance with the dialogue database 35 (S34). If the question content identification unit 52 is unable to identify the question content from the utterance information, the answer identifier determination unit 53 determines an answer identifier Q8 indicating that fact.

Next, the response unit 54 outputs the determined answer identifier to the interactive robot terminal 2 that output the utterance information (S35). The response unit 54 returns the answer identifier to the interactive robot terminal 2 that made the request as a result of analyzing the requested utterance information. At this point, the processing performed by the centralized processing device 3 ends. The processing performed by the interactive robot terminal 2 is resumed, and the interactive method is continuously executed in the interactive robot terminal 2.

In the interactive robot terminal 2, first, the reply receiving unit 43 receives the output reply identifier (S51). Next, the reply information determining unit 44 determines reply information corresponding to the output reply identifier according to the reply database 25 unique to each interactive robot terminal 2 (S52).
Here, in the preprocessing, the interactive robot terminal 2 identifies the language of the speech of the user U. The answer information determination unit 44 determines answer information that corresponds to the language identified by the data adjustment unit 41 and to the answer identifier received by the response reception unit 43.

At the start of practical use of the interactive robot system 1, answer information in at least a specific language (e.g., Japanese) is recorded in the answer database 25. This specific language is the language that is most frequently used and is also the language that is handled by the dialogue engine. Answer information in languages other than the specific language that the interactive robot terminal 2 can handle may be recorded at the start of practical use, or may be left blank. Here, it is assumed that answer information in all languages that can be handled is pre-recorded in the answer database 25 at the start of practical use (the case of starting practical use with a blank will be described later).

The answer information extracted from the answer database 25 is text data written in the same language as the spoken language. The data adjustment unit 41 obtains voice data converted from this text data (S53).
As in step S13, a known data conversion algorithm using a neural network is applied to convert the text data into voice data. The data conversion algorithm may be built into the terminal program P2 or may be provided by the external application 90.

Next, the speaker 27 outputs the converted voice data as answer information, and the display 28 outputs image information (including text) corresponding to the determined answer information (S54). In this way, an answer according to the meaning of the utterance is given to the user U in the same language as the utterance.
The above dialogue method will be described in more detail. For example, assume that a user U at station A utters in English, "Where is the washroom?" to the first dialogue robot terminal 2A. This utterance triggers the start of the dialogue method.

The first interactive robot terminal 2A identifies the language of the utterance as English based on the speech voice picked up by the microphone 26, and acquires the English text data "Where is the washroom?" indicating the content of the utterance. The acquired English text data is translated into Japanese text data. The first interactive robot terminal 2A outputs the translated Japanese text data "Where is the toilet?" to the centralized processing device 3 as speech information, and requests the centralized processing device 3 to analyze the speech information.

The centralized processing device 3 identifies the question as "where is the toilet" from the speech information, and determines the answer identifier "Q6" corresponding to the question according to the dialogue database 35. The centralized processing device 3 outputs the determined answer identifier "Q6" to the first interactive robot terminal 2A, which is the source of the inquiry.
The first interactive robot terminal 2A determines the English answer information corresponding to the answer identifier Q6 according to the first answer database 25A. The English text data extracted from the first answer database 25A is converted into voice data, and the answer information "The washroom is on your left" is spoken in English from the speaker 27. At the same time, the answer information is displayed in text on the display 28. A plan view of Station A showing the relative positions of the current location and the washroom may also be displayed.

The dialogue method ends with the output of this answer information. A conversation with an English speaker is established, and a guide service can be provided to the user U in an unmanned manner.
Here, for comparison with the present embodiment, a dialogue robot system according to a reference example will be described with reference to FIG. 5. In the dialogue robot system according to the reference example, a database referred to by a dialogue engine directly defines the correspondence between question contents and answer information. Since the answer information is specific to a station, the database and the dialogue engine including the database are provided in one-to-one correspondence with the dialogue robot terminal. A first dialogue robot terminal 902A installed at station A is communicatively connected to a first dialogue engine 903A including a first database 930A that records answer information specific to station A. A second dialogue robot terminal 902B installed at station B is communicatively connected to a second dialogue engine 903B including a second database 930B that records answer information specific to station B. Each dialogue engine 903A, 903B completes the process of determining the specific answer content by referring to the corresponding database 930A, 930B. The answer information determined by each of the dialogue engines 903A, 903B is output in the form of voice data from the dialogue engines 903A, 903B to the corresponding dialogue robot terminals 902A, 902B.

In contrast, the interactive robot system 1 according to this embodiment includes an interactive database 35 provided in the centralized processing device 3 and defining the correspondence between question contents and answer identifiers included in utterances, and an answer database 25 provided in the interactive robot terminal 2 and defining the correspondence between answer identifiers and answer information. When spoken to by a user U, the interactive robot terminal 2 transmits utterance information indicating the contents of the speech to the centralized processing device 3. The centralized processing device 3 identifies the question contents from the utterance information, and returns an answer identifier corresponding to the question contents to the interactive robot terminal 2. The interactive robot terminal 2 determines the answer information to be output to the user based on the answer identifier.

The centralized processing device 3 has the function of a "dialogue engine" that performs information processing such as analyzing utterance information, identifying the question content, and selecting an answer, but does not go so far as to determine the specific content of the answer. The specific content of the answer output to the user U is determined by the dialogue robot terminal 2. Two types of databases 25, 35 are provided separately for the centralized processing device 3 and the dialogue robot terminal 2. An answer identifier is involved in the process of determining the answer content from the question content, and the centralized processing device 3 and the dialogue robot terminal 2 share the execution of this process.

In the case where the dialogue engine completes the determination of the specific answer content as in the reference example, it becomes necessary to prepare a plurality of dialogue engines 903A, 903B for each of the plurality of interactive robot terminals 902A, 902B, and therefore the workload of constructing and maintaining the dialogue engines 930A, 930B becomes high.
In contrast, in the dialogue robot system 1 according to the present embodiment, information unique to the dialogue robot terminal 2 is recorded as answer information corresponding to an answer identifier in each of the multiple answer databases 25. This allows the dialogue engine to be shared by the multiple dialogue robot terminals 2, while outputting unique answer information to the user U. This makes it possible to reduce the workload of building and maintaining the dialogue engine.

Furthermore, during practical use of the interactive robot system 1, the content of questions may be added or changed in order to improve the guidance service. In the reference example, the content of questions needs to be added to all databases. In contrast, in this embodiment, it is sufficient to add the content of questions to only one dialogue database 25. This reduces the workload of the work required to improve the service.
The data amount of the answer identifier is about several bits, which is extremely small compared to the voice data. Since the answer identifier is transmitted from the centralized processing device 3 to the interactive robot terminal 2, the communication load can be reduced.

Each interactive robot terminal 2 inputs the contents of the user U's speech as voice data, and outputs text data converted from the input voice data to the centralized processing device 3 as speech information. Generally, text data has a smaller data volume than voice data. Therefore, the data volume of information to be transmitted from the interactive robot terminal 2 to the centralized processing device 3 is reduced. The centralized processing device 3 can identify the content of the question based on the text data, enabling rapid identification processing.

Each interactive robot terminal 2 is configured to determine answer information as text data based on the answer identifier, and outputs voice data converted from the text data as answer information. As a result, even if the answer information stored in the answer database 25 is text data, this text data is converted into voice data. Therefore, it is possible to realize a dialogue with the user U by vocalizing the answer information.

The answer database 25 stores answer information corresponding to answer identifiers for each of a plurality of languages, thereby enabling the interactive robot terminal 2 to output answer information in the same language as the utterance, and to realize an interaction with a user U who speaks various languages.
The interactive robot terminal 2 is preferably installed in a railway station. This allows for dialogue with the user U at the railway station. This allows for both labor saving for the railway operator's guidance work and enrichment of information provided to the user.

Although the embodiment of the present invention has been described above, the above configurations can be appropriately added to, modified, and/or deleted within the scope of the present invention.
For example, the system 1 may be put into practical use in a state where answer information in a language other than the specific language (e.g., Japanese) has not been entered in the answer database 25. When the utterance is in a language other than the specific language, the answer information determination unit 44 may not be able to determine the answer information even if it refers to the answer database 25 because the answer information corresponding to the determined answer identifier and the identified language is blank. In this case, the answer information determination unit 44 extracts answer information in the specific language corresponding to the same answer identifier. The data adjustment unit 41 machine-translates the extracted text data in the specific language into the spoken language, and records the translated text data in the answer database 25. In this way, an update process may be performed during practical use to fill in answer information in other languages in the answer database 25 as needed.

In the above embodiment, the identification process for identifying the language spoken by the user U is executed by applying an automatic language identification algorithm, but the user U may himself/herself select and input the language to be spoken.
In this case, the display 28 of each interactive robot terminal 2 is configured with a touch panel. When the interactive robot terminal 2 is in a standby state or when speech is detected in step S11, the controller 23 displays a button for selecting a language on the display 28. Before the user U starts speaking in earnest, the controller 23 may identify the language corresponding to the button touched by the user U as the user U's speech language.

The interactive robot system 1 can be applied to other customer service tasks in addition to railway operator guidance tasks, and may be introduced in hotels and hospitals.

1 Dialogue robot system 2 Dialogue robot terminal 3 Central processing device 25 Answer database 35 Dialogue database P Dialogue program P2 Terminal program P3 Engine program

Claims (8)

a plurality of dialogue robot terminals each outputting response information in response to a user's utterance;
A central processing device communicably connected to the plurality of interactive robot terminals;
a dialogue database provided in the centralized processing device and defining correspondence between question contents included in the utterance and answer identifiers;
a plurality of answer databases provided in each of the plurality of interactive robot terminals and defining a correspondence between the answer identifier and the answer information;
Equipped with
the interactive robot terminal outputs utterance information indicating the content of the utterance to the centralized processing device;
the centralized processing device identifies a question content from the speech information output from the interactive robot terminal, determines the answer identifier corresponding to the identified question content according to the interactive database, and outputs the determined answer identifier to the interactive robot terminal;
the interactive robot terminal determines answer information corresponding to the output answer identifier according to the answer database;
Conversational robot system. The plurality of interactive robot terminals each have the plurality of answer databases.
The interactive robot system according to claim 1 . Each of the interactive robot terminals receives an utterance content of a user as voice data, and converts the input voice data into text data and outputs the text data as the utterance information to the centralized processing device.
The interactive robot system according to claim 1 or 2. Each of the interactive robot terminals is configured to determine the answer information as text data based on the answer identifier, and output voice data converted from the text data as the answer information.
The interactive robot system according to any one of claims 1 to 3. the answer database stores the answer information corresponding to the answer identifier for each of a plurality of languages;
the interactive robot terminal outputs the response information in the same language as the utterance.
The interactive robot system according to any one of claims 1 to 4. The plurality of interactive robot terminals are installed at a plurality of railway stations, respectively.
The interactive robot system according to any one of claims 1 to 5. A dialogue method for use in a dialogue robot system including a plurality of dialogue robot terminals that output answer information in response to a user's utterance, and a central processing device communicatively connected to the plurality of dialogue robot terminals, comprising:
outputting utterance information indicating the content of the utterance from the interactive robot terminal to the centralized processing device;
identifying a question content from the output speech information in the centralized processing device;
determining, in the centralized processing device, the answer identifier corresponding to the identified question content according to a dialogue database that defines a correspondence between question content included in the utterance and an answer identifier;
outputting the determined answer identifier from the centralized processing device to the interactive robot terminal;
determining, in the interactive robot terminal, the answer information corresponding to the output answer identifier according to an answer database that defines a correspondence between the answer identifier and the answer information;
The method of interaction comprises: A dialogue program for causing a computer to execute a dialogue method used in a dialogue robot system including a plurality of dialogue robot terminals that output answer information in response to a user's utterance, and a central processing device communicatively connected to the plurality of dialogue robot terminals,
outputting utterance information indicating the content of the utterance from the interactive robot terminal to the centralized processing device;
identifying a question content from the output speech information in the centralized processing device;
determining, in the centralized processing device, the answer identifier corresponding to the identified question content according to a dialogue database that defines a correspondence between question content included in the utterance and an answer identifier;
outputting the determined answer identifier from the centralized processing device to the interactive robot terminal;
determining, in the interactive robot terminal, the answer information corresponding to the output answer identifier according to an answer database that defines a correspondence between the answer identifier and the answer information;
An interactive program comprising:

Images