JP2022043901A - Dialogue system, interactive robot, program, and information processing method - Google Patents

Abstract

To improve the accuracy of estimating a subject of conversation in a dialogue with a user.SOLUTION: Processors (11, 21) included in a dialogue system 1 execute: estimation processing (S103) of estimating a subject of conversation based on a first user voice uttered by a user; generation processing (S105) of generating a response voice to respond to the first user voice; and determination processing (S109-S111) of, based on whether or not a second user voice uttered by the user in correspondence to an output of the response voice indicates negative content, determines whether or not the subject of conversation estimated by the estimation processing is correct.SELECTED DRAWING: Figure 2

Description

The present invention relates to a technique for interacting with a user.

Techniques for interacting with users are known. For example, Patent Document 1 describes a technique for estimating a topic corresponding to an input voice by using a plurality of language models hierarchically configured according to the type of topic. The technique estimates the topic by selecting one language model based on the tentative recognition result of the input speech, the similarity between each language model, the reliability of the recognition result, and the depth of the hierarchy.

Patent No. 5212910 (published March 8, 2013)

However, in the technique described in Patent Document 1, if the tentative recognition result is incorrect, the accuracy of topic estimation becomes low. Therefore, the technique may estimate a topic different from the actual topic, and there is room for improvement in the estimation accuracy of the topic.

One aspect of the present invention is to realize a technique for improving the estimation accuracy of a topic in a dialogue with a user.

In order to solve the above problems, the dialogue system according to one aspect of the present invention includes one or more processors. The one or more processors execute the estimation process, the generation process, and the determination process. Further, the information processing method according to one aspect of the present invention is an information processing method executed by one or a plurality of processors. The information processing method includes an estimation step, a generation step, and a judgment step.

In the estimation process (estimation step), the one or more processors estimate the topic based on the first user voice spoken by the user. In the generation process (generation step), the one or more processors generate a response voice in response to the first user voice. In the determination process (determination step), the one or more processors are based on whether or not the second user voice spoken by the user in response to the output of the response voice shows negative content. It is determined whether or not the topic estimated by the estimation process is correct.

According to one aspect of the present invention, it is possible to realize a technique for improving the estimation accuracy of a topic in a dialogue with a user.

It is a block diagram which shows the structure of the dialogue system which concerns on one Embodiment of this invention. It is a flowchart which shows the flow of the information processing method which concerns on one Embodiment of this invention. It is a flowchart which shows the detailed flow of the estimation process included in the information processing method shown in FIG. It is a figure which shows the specific example of the similar keyword database in one Embodiment of this invention. It is a figure explaining the application example of the dialogue system which concerns on one Embodiment of this invention.

Hereinafter, an embodiment of the present invention will be described with reference to the drawings.

<Overview of dialogue system>
The dialogue system 1 is a system that interacts with a user by acquiring a user voice spoken by the user and generating and outputting a response voice that responds to the acquired user voice. The dialogue system 1 estimates the topic based on the first user voice. Further, the dialogue system 1 determines whether the estimated topic is correct or incorrect based on whether or not the second user voice shows negative content. The second user voice is a voice spoken by the user in response to the output of the response voice in response to the first user voice. In the present embodiment, the target user of the dialogue system 1 is a care recipient, an elderly person, or the like. The dialogue system 1 is used for interacting with these users. However, the users targeted by the dialogue system 1 are not limited to these examples.

<Configuration of Dialogue System 1>
The configuration of the dialogue system 1 will be described with reference to FIG. FIG. 1 is a block diagram showing a configuration of a dialogue system 1 according to an embodiment of the present invention. As shown in FIG. 1, the dialogue system 1 includes a dialogue robot 10 and a server 20.

(Configuration of dialogue robot 10)
The configuration of the dialogue robot 10 will be described with reference to FIG. As shown in FIG. 1, the dialogue robot 10 includes a controller 110, a microphone 120, and a speaker 130. For example, when the dialogue robot 10 has a face (not shown), the microphone 120 and the speaker 130 may be included in the face.

The controller 110 controls the operation of the entire dialogue robot 10. The controller 110 includes a processor 11, a primary memory 12, a secondary memory 13, a communication interface 14, and an input / output interface 15. The processor 11, the primary memory 12, the secondary memory 13, the communication interface 14, and the input / output interface 15 are connected to each other via a bus.

The program P1 is stored in the secondary memory 13. The program P1 is a program for causing the processor 11 to execute at least a part of the information processing method S described later. The processor 11 expands the program P1 stored in the secondary memory 13 on the primary memory 12. Then, the processor 11 executes each step included in the information processing method S according to the instruction included in the program P1 expanded on the primary memory 12.

Devices that can be used as the processor 11 include, for example, a CPU (Central Processing Unit), a GPU (Graphic Processing Unit), a DSP (Digital Signal Processor), an MPU (Micro Processing Unit), an FPU (Floating point number Processing Unit), and a PPU. (Physics Processing Unit), a microcontroller, or a combination thereof can be mentioned. The processor 11 is sometimes called an "arithmetic unit".

Further, as a device that can be used as the primary memory 12, for example, a semiconductor RAM (Random Access Memory) can be mentioned. The primary memory 12 is sometimes referred to as a "main storage device". Devices that can be used as the secondary memory 13 include, for example, a flash memory, an HDD (Hard Disk Drive), an SSD (Solid State Drive), an ODD (Optical Disk Drive), and an FDD (Floppy (registered trademark) Disk Drive). , Or a combination of these. The secondary memory 13 is sometimes called an "auxiliary storage device". The secondary memory 13 may be built in the controller 110, or may be another computer (for example, the server 20) connected to the controller 110 (dialogue robot 10) via the communication interface 14 or the input / output interface 15. It may be built in. In the present embodiment, the storage in the controller 110 is realized by two memories (primary memory 12 and secondary memory 13), but the present invention is not limited to this. That is, the memory in the controller 110 may be realized by one memory. In this case, for example, a certain storage area of the memory may be used as the primary memory 12, and the other storage area of the memory may be used as the secondary memory 13.

Another computer is connected to the communication interface 14 by wire or wirelessly via the network N1. Here, at least the server 20 is connected as another computer. Examples of the communication interface 14 include interfaces such as Ethernet (registered trademark) and Wi-Fi (registered trademark). Available networks include PAN (Personal Area Network), LAN (Local Area Network), CAN (Campus Area Network), MAN (Metropolitan Area Network), WAN (Wide Area Network), GAN (Global Area Network), or , Internetworks including these networks. Internetwork may be an intranet, an extranet, or the Internet.

A microphone 120 and a speaker 130 are connected to the input / output interface 15. Examples of the input / output interface 15 include interfaces such as USB (Universal Serial Bus), ATA (Advanced Technology Attachment), SCSI (Small Computer System Interface), and PCI (Peripheral Component Interconnect).

(Configuration of server 20)
The configuration of the server 20 will be described with reference to FIG. As shown in FIG. 1, the server 20 includes a processor 21, a primary memory 22, a secondary memory 23, and a communication interface 24. The processor 21, the primary memory 22, the secondary memory 23, and the communication interface 24 are connected to each other via a bus.

The secondary memory 23 stores a program P2, a plurality of topic keyword databases (DB) 231, a plurality of similar keyword databases (DB) 232, and a voice database (DB) 233. Details of these DBs 231 to 233 will be described later. The program P2 is a program for causing the processor 21 to execute at least a part of the information processing method S described later. The processor 21 expands the program P2 stored in the secondary memory 23 on the primary memory 22. Then, the processor 21 executes each step included in the information processing method S according to the instruction included in the program P2 expanded on the primary memory 22.

Devices that can be used as the processor 21 include, for example, a CPU (Central Processing Unit), a GPU (Graphic Processing Unit), a DSP (Digital Signal Processor), an MPU (Micro Processing Unit), an FPU (Floating point number Processing Unit), and a PPU. (Physics Processing Unit), a microcontroller, or a combination thereof can be mentioned. The processor 21 is sometimes called an "arithmetic unit".

Further, as a device that can be used as the primary memory 22, for example, a semiconductor RAM (Random Access Memory) can be mentioned. The primary memory 22 is sometimes referred to as a "main storage device". Devices that can be used as the secondary memory 23 include, for example, a flash memory, an HDD (Hard Disk Drive), an SSD (Solid State Drive), an ODD (Optical Disk Drive), and an FDD (Floppy (registered trademark) Disk Drive). , Or a combination of these. The secondary memory 23 is sometimes called an "auxiliary storage device". The secondary memory 23 may be built in the server 20, or may be built in another computer (for example, a computer constituting the cloud server) connected to the server 20 via the communication interface 24. good. In the present embodiment, the storage in the server 20 is realized by two memories (primary memory 22 and secondary memory 23), but the storage is not limited to this. That is, the storage in the server 20 may be realized by one memory. In this case, for example, a certain storage area of the memory may be used as the primary memory 22, and the other storage area of the memory may be used as the secondary memory 23.

Another computer is connected to the communication interface 24 by wire or wirelessly via the network N1. Here, at least the dialogue robot 10 is connected as another computer. Examples of the communication interface 24 include interfaces such as Ethernet (registered trademark) and Wi-Fi (registered trademark). Available networks include PAN (Personal Area Network), LAN (Local Area Network), CAN (Campus Area Network), MAN (Metropolitan Area Network), WAN (Wide Area Network), GAN (Global Area Network), or , Internetworks including these networks. Internetwork may be an intranet, an extranet, or the Internet.

<Flow of information processing method S>
The information processing method S executed by the dialogue system 1 will be described with reference to FIG. FIG. 2 is a flowchart showing the flow of the information processing method S. In FIG. 2, the left figure shows the process executed by the processor 11 (dialogue robot 10), and the right figure shows the process executed by the processor 21 (server 20). As shown in FIG. 2, the information processing method S includes steps S101 to S114.

(Step S101)
In step S101, the processor 11 of the dialogue robot 10 executes a voice acquisition process of acquiring the voice input via the microphone 120 as the first user voice. Further, the processor 11 temporarily stores the acquired first user voice in the primary memory 12 or the secondary memory 13.

(Step S102)
In step S102, the processor 11 executes voice recognition processing for the first user voice. The voice recognition result by the voice recognition process is text information corresponding to the first user voice. As the speech recognition process, a known technique can be applied. For example, the processor 11 may read and execute a known voice recognition program stored in the secondary memory 13, or may use a voice recognition service provided by a cloud server (not shown).

(Steps S103 to S104)
In step S103, the processor 11 executes an estimation process for estimating a topic based on the first user voice. The estimation process is executed using the voice recognition result in step S102. By executing the step, the keyword included in the first user voice is specified, and the topic related to the first user voice is estimated. The processor 11 makes an inquiry to the server 20 in order to execute step S103. In step S104, the server 20 responds to an inquiry by searching for a plurality of topic keyword DB 231 and a plurality of similar keyword DB 232. Details of steps S103 to S104 will be described later.

(Step S105)
In step S105, the processor 11 executes a generation process for generating a response voice in response to the first user voice. Specifically, the processor 11 generates text information for a response by using the text information which is the voice recognition result of the first user voice. Further, the processor 11 generates a response voice from the text information for the response by using the speech synthesis technique.

(Specific example of response voice)
Here, a specific example of the response voice will be described. For example, the processor 11 generates a response voice including a repeat of a keyword included in the first user voice, or a response voice including a question related to the keyword. Repeating is to repeat the keyword. Further, it is desirable that the question related to the keyword is a question that delves into the content of the first user voice.

For example, it is assumed that the text information "Today is good weather" is obtained as the voice recognition result of the first user voice. Further, it is assumed that "good" and "weather" are specified as keywords from the voice recognition result. In this case, the processor 11 generates a response voice "good weather" including the repetition of these keywords.

Further, for example, it is assumed that the text information "I went to the park yesterday" is obtained as the first user voice recognition result. Further, it is assumed that "park" and "went" are specified as keywords from the voice recognition result. In this case, the processor 11 generates a response voice "where did you go to the park" containing questions related to these keywords.

(Step S106)
In step S106, the processor 11 executes a voice output process for outputting the response voice from the speaker 130.

(Step S107)
In step S107, the processor 11 executes a voice acquisition process of acquiring the voice input via the microphone 120 as a second user voice.

(Step S108)
In step S108, the processor 11 executes voice recognition processing for the second user voice. The voice recognition result by the voice recognition process is text information corresponding to the second user voice.

(Step S109)
In step S109, the processor 11 determines whether the second user voice shows negative content.

(Specific example 1 of negative judgment processing)
For example, the processor 11 executes a negative determination process of whether or not the second user voice shows a negative content based on the above-mentioned voice recognition result. Specifically, the processor 11 determines that the second user voice indicates a negative content when the voice recognition result of the second user voice contains a negative keyword. Examples of negative keywords include, but are not limited to, "no,""no," and "not." In this case, for example, the secondary memory 13 stores a negative keyword in advance. The processor 11 refers to the voice recognition result of the second user voice and the negative keyword stored in the secondary memory 13, and performs the negative determination process.

(Specific example 2 of negative judgment processing)
Further, the processor 11 may execute the above-mentioned negative determination process in place of or in addition to the method based on the voice recognition result described in the first embodiment by using another method. Specific example 2 is an example of another method, and is a method of performing a negative judgment process based on a user's gesture or facial expression. For example, negative gestures include shaking the head. In addition, negative facial expressions include a lowered corner of the mouth. For example, the processor 11 uses a camera (not shown) to acquire an image of a user speaking a second user voice. In addition, the processor 11 analyzes the acquired image to determine whether or not the image exhibits negative gesture features or negative facial expression features. In this case, for example, the secondary memory 13 stores in advance negative gesture features or facial expression features in the image. The processor 11 performs the negative determination process with reference to an image of a user speaking a second user voice and a feature stored in the secondary memory 13.

(Step S110)
In the case of Yes in step S109, in step S110, the processor 11 determines that the topic related to the first user voice is another topic. The other topic is a topic other than the topic estimated in step S103. In other words, the processor 11 determines that the topic estimated in step S103 is incorrect. For example, at this time, the processor 11 may output a response voice indicating that the estimated topic was incorrect. Specific examples of such response voices include, but are not limited to, "I'm sorry, I made a mistake."

When the processor 11 executes step S110, the processor 11 ends the information processing method S. The processor 11 erases the temporarily stored first user voice from the primary memory 12 or the secondary memory 13 before the information processing method S is terminated. Further, for example, the processor 11 may repeat the information processing method S again after finishing the information processing method S.

(Step S111)
If No in step S109, in step S111, the processor 11 determines that the topic estimated in step S103 is correct.

(Step S112)
In step S112, the processor 11 determines whether or not the estimated topic satisfies a predetermined condition. In the present embodiment, the predetermined condition is a condition that "the topic is related to health".

If No in step S112, the information processing method S ends. The processor 11 erases the temporarily stored first user voice from the primary memory 12 or the secondary memory 13 before the information processing method S is terminated. Further, for example, the processor 11 may repeat the information processing method S again after finishing the information processing method S.

(Steps S113 to S114)
In the case of Yes in step S112, in step S113, the processor 11 requests the recording of the first user voice by transmitting the first user voice stored in the primary to the server 20. In step S114, the processor 21 of the server 20 executes a recording process of recording the received first user voice in the voice DB 233. The processor 21 transmits the access information for the first user voice recorded in the voice DB 233 to the dialogue robot 10. The first user voice recorded in the voice DB 233 can be downloaded, played back, or both by being accessed using the access information. Upon receiving the access information, the processor 11 erases the temporarily stored first user voice from the primary memory 12 or the secondary memory 13.

(Voice DB 233)
The voice DB 233 stores the first user voice. Further, the voice DB 233 may store related information in association with the first user voice. Examples of related information include the date and time, keywords included in the first user voice, topics estimated by estimation processing, user identification information, and the current position of the dialogue robot 10. In this case, in step S113, the processor 11 of the dialogue robot 10 transmits the related information to the server 20 together with the first user voice. The processor 21 of the server 20 associates these information received from the dialogue robot 10 and records it in the voice DB.

(Step S115)
In step S115, the processor 11 executes a transmission process for transmitting information about the user to the outside. The information about the user includes access information for the first user voice. Further, the information about the user may include the related information stored in the voice DB in association with the first user voice. An example of a destination for sending information about a user is, but is not limited to, an administrator (family, caregiver, attending physician, etc.) who manages the health of the user. Further, an example of a means for transmitting information about a user is, but is not limited to, e-mail. For example, the secondary memory 13 stores information indicating a destination and a transmission means in advance.

(Details of topic estimation processing)
Next, the details of the topic estimation process in steps S103 to S104 will be described with reference to FIG. FIG. 3 is a flowchart showing a detailed flow of the topic estimation process. As shown in FIG. 3, the topic estimation process includes steps S201 to S206.

(Step S201)
In step S201, the processor 11 of the dialogue robot 10 extracts one or a plurality of keywords from the voice recognition result of the first user voice. For example, a known natural language processing technique (for example, morphological analysis or the like) can be applied to the keyword extraction process.

(Step S202)
In step S202, the processor 11 estimates the topic based on which of the plurality of topic keyword DB 231 each extracted keyword is included in. In order to execute the process of the step, the processor 11 inquires the server 20 about the topic keyword DB 231 including each keyword, and the server 20 responds to the inquiry.

(Topic keyword DB231)
Here, the topic keyword DB 231 will be described. The secondary memory 23 of the server 20 stores the topic keyword DB 231 in association with each of the plurality of topics. Each topic keyword DB 231 includes one or more keywords used in the topic. The same keyword may be included in a plurality of topic keyword DB231. Multiple topics include "health-related topics". "Health-related topics" include, for example, topics related to physical condition or illness. In addition, the plurality of topics may include, but are not limited to, for example, "a topic reminiscent of a special fraud such as oleore fraud" and "a topic related to human relations". ..

(When there is one topic keyword DB231 including keywords)
When one or a plurality of extracted keywords are included in one topic keyword DB 231, the processor 11 estimates the topic associated with the topic keyword DB 231 as a topic related to the first user voice. As such a case, there is a case where the extracted keyword is one and the topic keyword DB 231 including the keyword is one. Further, as such a case, a plurality of keywords may be extracted and the topic keyword DB 231 including each keyword may be all the same.

(When there are multiple topic keyword DB231 including keywords)
Further, when the extracted one or a plurality of keywords are included in the plurality of topic keyword DB 231s, the processor 11 estimates the topic associated with any of them as the topic related to the first user voice. As such a case, there is a case where the extracted keyword is one and the keyword is included in a plurality of topic keyword DB231. Further, as such a case, a plurality of keywords may be extracted, and at least two of them may be included in the topic keyword DB 231 which is different from each other.

For example, the processor 11 selects one of the topics associated with the corresponding plurality of topic keyword DBs based on a predetermined rule. The processor 11 estimates the selected topic as a topic related to the first user voice. Specific examples of predetermined rules include, but are not limited to, (1) fixed priority of topics, (2) dynamic priority of topics, and (3) number of keywords.

(1) When based on a fixed priority of a topic, the secondary memory 13 stores in advance a fixed priority determined between a plurality of topics. The processor 11 selects the topic having the highest fixed priority from the topics associated with the corresponding plurality of topic keyword DBs.

(2) Based on the dynamic priority of topics, the secondary memory 13 stores the history of topics determined to be correct in step S111 executed in the past. The processor 11 dynamically changes the priority of the topic according to the history of the topic. The processor 11 selects the topic having the highest dynamic priority among the topics associated with the corresponding plurality of topic keywords DB 231. For example, the processor 11 may have the highest priority of the topic determined to be correct in the latest step S111. Further, the processor 11 may raise the priority of the topic in descending order of the number of times determined to be correct in the latest predetermined number of times or the number of times determined to be correct in step S111 during the latest predetermined number of times.

(3) Based on the number of keywords, the processor 11 selects the one containing the most extracted keywords from the corresponding plurality of topic keyword DB231.

(Step S203)
In step S203, the processor 11 determines whether or not the topic has been estimated in step S202. For example, if there is no topic keyword DB 231 including one or a plurality of extracted keywords, the processor 11 determines that the topic has not been estimated.

(In the case of Yes in step S203)
If Yes in step S203, the processor 11 ends the topic estimation process in step S103. As a result, one or a plurality of keywords extracted in step S201 are applied as the keywords specified in step S103. Further, as the topic estimated in step S103, the topic estimated in step S202 is applied.

(When No in step S203)
If No in step S203, the processor 11 executes the process from step S204.

(Step S204)
In step S204, the processor 11 determines whether or not each of the extracted keywords is a similar keyword. In order to execute the process of the step, the processor 11 inquires the server 20 of the topic keyword DB 231 including each keyword as a similar keyword, and the server 20 responds to the inquiry.

(Similar keyword DB232)
Here, the similar keyword DB232 will be described. The secondary memory 23 of the server 20 stores the similar keyword DB 232 in association with each of the plurality of topics. Each similar keyword DB232 stores the correct keyword used in the topic and the similar keyword similar to the correct keyword in association with each other. Similar keywords are keywords that are likely to be erroneously recognized by voice recognition for the user's voice that utters the correct keyword.

FIG. 4 is a diagram showing a specific example of the similar keyword DB 232. As shown in FIG. 4, similar keywords "Ittai (one)", "Itta (go)", "Ita (was)", and "Thai" are associated with the correct keyword "pain", respectively. There is. For example, these similar keywords are different from "painful" obtained by voice recognition processing for the user voice uttered "painful".

The correct answer keyword "painful" shown in FIG. 4 is, for example, a keyword used in "a topic related to health" and is included in the topic keyword DB 231 associated with the topic.

(If No in step S204)
If No in step S204, the processor 11 executes step S110 in FIG. 2 and determines that the topic related to the first user voice is another topic. The other topic here is a topic that is neither of the topics associated with the plurality of topic keywords DB231. In this case, the processor 11 may end the information processing method S, and then, for example, execute the information processing method S again.

(In the case of Yes in step S204)
If Yes in step S204, the processor 11 executes the process of step S205.

(Step S205)
In step S205, the processor 11 replaces the similar keyword included in the voice recognition result of the first user voice with the correct answer keyword associated with the similar keyword in the similar keyword DB232. That is, the processor 11 replaces a similar keyword with a high possibility of erroneous recognition with a correct answer keyword among one or a plurality of keywords extracted from the voice recognition result of the first user voice.

In step S204, it may be determined that a certain extracted keyword is included in a plurality of similar keyword DB232 as a similar keyword. In this case, the processor 11 selects one of the plurality of applicable similar keyword DB 232s based on a predetermined rule. Further, the processor 11 may execute the above-mentioned replacement process using the selected similar keyword DB232. As a specific example of the rule for selecting one from a plurality of similar keyword DB232, as in the predetermined rule described in step S202, (1) a fixed priority of the topic and (2) a dynamic topic. Priority and (3) the number of keywords can be mentioned, but are not limited to these.

For example, it is assumed that two keywords "head" and "integral" are extracted from the voice recognition result of the first user voice. Here, of the two keywords, "one" is a similar keyword. In addition, the correct keyword "painful" is associated with the similar keyword "one". Therefore, the processor 11 replaces the similar keyword "integral" included in the speech recognition result with the correct answer keyword "painful". As a result, the speech recognition result after the replacement includes two keywords "head" and "painful".

(Step S206)
In step S206, the processor 11 executes a process of estimating a topic based on the voice recognition result after the replacement. The process of estimating the topic is the same as the process of step S202. However, the difference is that each keyword included in the voice recognition result after replacement is used instead of the voice recognition result before replacement.

When the process of step S206 is executed, the processor 11 ends the topic estimation process in step S103. As a result, as the keyword specified in step S103, one or a plurality of keywords included in the replaced voice recognition result are applied. Further, as the topic estimated in step S103, the topic estimated in step S206 is applied.

After that, in step S105 of FIG. 2, the processor 11 generates a response voice based on the voice recognition result after the replacement. For example, in the above example, the speech recognition result after the replacement includes two keywords "head" and "painful". In this case, in step S105, the processor 11 generates a response voice "Do you have a headache?" That repeats the keyword.

As described above, in the topic estimation process in step S103 of FIG. 2, the processor 11 estimates any of a plurality of topics.

This completes the description of the flow of the information processing method S executed by the dialogue system 1.

<Application example>
An application example of the dialogue system 1 will be described with reference to FIG. FIG. 5 is a diagram illustrating an application example of the dialogue system 1. The dialogue robot 10 shown in FIG. 5 is arranged beside the care recipient for the purpose of having a dialogue with the care recipient. As shown in FIG. 5, the dialogue robot 10, the care recipient, and the caregiver perform the following steps N1 to N7.

(Step N1: Utterance by the care recipient)
The care recipient says, "I have a headache."

(Step N2: Estimating the topic by the dialogue robot 10)
The dialogue robot 10 acquires the voice spoken by the care recipient as the first user voice. Further, the dialogue robot 10 performs a replacement process using the similar keyword DB 232 for the voice recognition result "head is one" of the first user voice. As a result, the speech recognition result after the replacement includes the keywords "head" and "painful". Further, it is assumed that the keywords "head" and "painful" are included in the topic keyword DB 231 associated with the "topic related to health". Therefore, the dialogue robot 10 estimates "health-related topics" (steps S101 to S104 in FIG. 2).

(Step N3: Response by the dialogue robot 10)
The dialogue robot 10 uses the keywords "head" and "painful" included in the replaced voice recognition result to generate and output a response voice "does your head hurt?" That repeats these keywords (does it hurt your head? " Steps S105 to S106).

(Step N4: Negative response by the care recipient)
A case where the care recipient gives a negative response such as “No” to the response voice of the dialogue robot 10 will be described. In this case, the dialogue robot 10 acquires the negative response of the care recipient as the second user voice. Further, the dialogue robot 10 determines that the topic of the care recipient is other than the "topic related to health" because the second user voice shows a negative content. Further, the dialogue robot 10 outputs a voice such as "I'm sorry, I made a mistake" (steps S107 to S109, S110). Subsequently, the operation from step N1 is repeated.

(Step N5: Acknowledgment by the care recipient)
A case where the care recipient gives an affirmative response such as “yes” to the response voice of the dialogue robot 10 will be described. The dialogue robot 10 acquires the affirmative response of the care recipient as a second user voice. Further, the dialogue robot 10 determines that the estimated "health-related topic" is correct because the second user voice does not show a negative content (steps S107 to S109, S111).

(Step N6: Notification from the dialogue robot 10 to the caregiver)
Next, since the topic of the care recipient is related to health, the dialogue robot 10 records the first user voice in the voice DB 233 by transmitting it to the server 20. Further, the dialogue robot 10 stores the caregiver's e-mail address as the contact information of the care recipient. Therefore, the dialogue robot 10 sends an e-mail containing the information of the care recipient to the e-mail address of the caregiver. The transmitted e-mail includes access information for the first user voice recorded in the voice DB 233 (steps S112 to S115).

(Step N7: Playback of the first user voice by the caregiver)
The caregiver accesses the voice DB 233 of the server 20 by using the access information included in the received e-mail, and reproduces the first user voice “head hurts” of the care recipient.

<Effect of this embodiment>
The dialogue system 1 according to the present embodiment determines whether or not the topic estimated based on the voice recognition result of the first user voice is correct, and whether or not the second user voice shows negative content. to decide. As a result, the estimation accuracy of the topic is improved as compared with the case where the second user voice is not taken into consideration.

Further, the dialogue system 1 according to the present embodiment replaces the similar keyword included in the voice recognition result of the first user voice with the correct answer keyword associated with the similar keyword, and based on the voice recognition result after the replacement. Estimate the topic. As a result, the possibility of erroneously recognizing the first user voice can be reduced, so that the estimation accuracy of the topic is further improved.

Further, in the dialogue system 1 according to the present embodiment, when the estimated topic is related to health, the first user voice is recorded in the voice DB 233, and the recorded access information to the first user voice is recorded by the user administrator. Send to your contacts. As a result, the dialogue robot 10 can detect health-related utterances that are likely to have an urgent need and promptly notify the detected utterances to the outside while being used for the purpose of enjoying the dialogue by the user.

[Modification example]
(Variation example of the number of topics)
In the present embodiment, the dialogue system 1 has been described as presuming any one of a plurality of topics as a topic related to the first user voice. Not limited to this, the dialogue system 1 may estimate whether or not the topic related to the first user voice is one specific topic. In this case, the server 20 stores the topic keyword DB 231 and the similar keyword DB 232 related to a specific topic one by one. One particular topic may be, for example, a topic related to health. In this case, the dialogue system 1 can accurately estimate whether or not the topic related to the first user voice is a "topic related to health".

(Modification example of predetermined conditions)
Further, in the present embodiment, an example in which the condition that "the topic is related to health" is applied as a predetermined condition for determining whether or not to transmit information to the outside has been described. However, the predetermined condition is not limited to this, and may be a condition related to other topics.

(Modified example of topic estimation processing)
Further, in the topic estimation process of the present embodiment, the processor 11 may omit steps S202 and S203 of FIG. 3 depending on the accuracy of the voice recognition result. For example, the processor 11 may execute steps S202 and S203 when the accuracy of the speech recognition result is equal to or greater than the threshold value and may be omitted when the accuracy is less than the threshold value. As a result, when the accuracy of the speech recognition result is high, the processor 11 first estimates the topic without performing the replacement process. Therefore, even though the accuracy is high, the possibility that an erroneous topic is presumed by performing the replacement process is reduced. Further, as a result, when the accuracy of the voice recognition result is low, the processor 11 first executes the replacement process and then estimates the topic. Therefore, the possibility that an erroneous topic is estimated using the speech recognition result with low accuracy is reduced.

(Modified example in which the server 20 executes the main steps)
Further, in the information processing method S according to the present embodiment, the server 20 may execute a part of the steps executed by the dialogue robot 10. For example, the dialogue robot 10 may execute the voice acquisition process (steps S101 and S107) and the voice output process (step S106), and the server 20 may execute other steps. In this case, the dialogue robot 10 may transmit the acquired user voice to the server 20, receive the response voice from the server 20, and output it.

(Modification example of storage location of DB231 to 233)
Further, in the present embodiment, the topic keyword DB 231 and the similar keyword DB 232 and the voice DB 233 have been described as being stored in the secondary memory 23 of the server 20. Not limited to this, a part or all of these DBs 231 to 233 may be stored outside the dialogue system 1 (for example, a cloud server or the like). Further, a part or all of these DBs 231 to 233 may be stored in the secondary memory 13 of the dialogue robot 10.

(Modification example in which the dialogue robot 10 executes all steps)
Further, when the server 20 does not store any of the above-mentioned DBs 231 to 233, the dialogue robot 10 may execute all the steps of the information processing method S according to the present embodiment. In this case, the server 20 can be omitted. In other words, in the present embodiment, a configuration is adopted in which a plurality of processors (processor 11 of the interactive robot 10 and processor 21 of the server 20) cooperate to execute the information processing method S. The present invention is not limited to this, and a configuration in which the information processing method S is executed using a single processor 11 of the interactive robot 10 may be adopted.

(A configuration that replaces the dialogue robot 10)
Further, in the present embodiment, the dialogue system 1 may include a computer having a processor and a memory such as a smartphone, a tablet, a smart speaker, a personal computer, etc., instead of the dialogue robot 10. In this case, the program P1 similar to that of the dialogue robot 10 is stored in the memory, and the processor reads and executes the program P1. As a result, the dialogue system 1 operates in the same manner as in the above-described embodiment and has the same effect.

〔summary〕
The dialogue system according to aspect 1 comprises one or more processors. The one or more processors execute the estimation process, the generation process, and the determination process. The estimation process is a process of estimating a topic based on the first user voice spoken by the user. The generation process is a process of generating a response voice in response to the first user voice. In the judgment process, whether or not the topic estimated by the estimation process is correct is based on whether or not the second user voice spoken by the user in response to the output of the response voice shows negative content. It is a process to judge.

With the above configuration, it is determined whether or not the topic estimated based on the first user voice is correct based on whether or not the second user voice shows negative content. The second user voice is a voice generated in response to the first user voice. As a result, the estimation accuracy of the topic is improved as compared with the case where the second user voice is not taken into consideration.

The dialogue system according to the second aspect has the following features in addition to the features of the dialogue system according to the first aspect. That is, in the dialogue system according to the second aspect, the one or more processors refer to a similar keyword database in which a correct keyword used in a specific topic and a similar keyword similar to the correct keyword are associated with each other. Further, the one or a plurality of processors further executes a replacement process of replacing the similar keyword included in the voice recognition result of the first user voice with a correct answer keyword associated with the similar keyword in the similar keyword database. do. Further, the one or more processors execute the estimation process and the generation process based on the voice recognition result after the replacement.

With the above configuration, it is less likely that the voice recognition result after replacement is incorrect. As a result, the estimation accuracy of the topic is further improved based on the speech recognition result after such replacement.

The dialogue system according to the third aspect has the following features in addition to the features of the dialogue system according to the second aspect. That is, in the dialogue system according to the third aspect, the one or the plurality of processors refer to the similar keyword database associated with each of the plurality of topics including the specific topic, and in the estimation process, the plurality of topics. Estimate one of them.

With the above configuration, it is possible to reduce the possibility that the voice recognition result of the first user voice is erroneous for each of the plurality of topics.

The dialogue system according to the fourth aspect has the following features in addition to the features of the dialogue system according to any one of the first to third aspects. That is, in the dialogue system according to the fourth aspect, the one or more processors further execute a transmission process of transmitting information about the user to the outside when the topic determined to be correct by the determination process satisfies a predetermined condition. ..

With the above configuration, it is possible to detect a user's utterance regarding a topic satisfying a predetermined condition and promptly notify the outside.

The dialogue system according to the fifth aspect has the following features in addition to the features of the dialogue system according to the fourth aspect. That is, in the dialogue system according to the fifth aspect, the one or more processors further execute a recording process for recording the first user voice in the memory. Further, in the transmission process, the one or more processors include the access information for the first user voice recorded in the memory in the information about the user and transmits the information.

With the above configuration, the receiver of the access information can access and play back the first user voice related to the topic satisfying the predetermined condition.

The dialogue system according to the sixth aspect has the following features in addition to the features of the dialogue system according to any one of the first to fifth aspects. That is, in the dialogue system according to the sixth aspect, the one or more processors estimate a topic related to health as the topic in the estimation process.

With the above configuration, it is possible to more accurately estimate whether or not the user's topic is related to health.

The dialogue robot according to the seventh aspect is the dialogue robot included in the dialogue system according to any one of the first to sixth aspects, and includes any one or the plurality of processors. The processor included in the dialogue robot executes at least a voice acquisition process for acquiring the first user voice and the second user voice, and a voice output process for outputting the response voice.

With the above configuration, the user can use the dialogue system according to any one of aspects 1 to 6 in the form of a dialogue robot that interacts with the user.

The program according to the eighth aspect is a program for operating the dialogue system of any one of the first aspect to the sixth aspect, and causes the one or a plurality of processors to execute each of the above processes.

With the above configuration, the same effect as that of the dialogue system according to the first aspect is obtained.

The information processing method according to the ninth aspect includes an estimation step, a generation step, and a determination step. In the estimation step, one or more processors estimate the topic based on the first user voice spoken by the user. In the generation step, the one or more processors generate a response voice in response to the first user voice. In the determination step, the one or more processors are based on the estimation step based on whether or not the second user voice spoken by the user in response to the output of the response voice shows negative content. Determine if the estimated topic is correct.

With the above configuration, the same effect as that of the dialogue system according to the first aspect is obtained.

1 Dialogue system 10 Dialogue robot 11, 21 Processor 12, 22 Primary memory 13, 23 Secondary memory 14, 24 Communication interface 15 Input / output interface 20 Server 110 Controller 120 Microphone 130 Speaker 231 Topic keyword DB
232 Similar keyword DB
233 Voice DB

Claims (9)

An interactive system with one or more processors
The one or more processors
Estimating processing that estimates the topic based on the first user voice spoken by the user,
A generation process for generating a response voice in response to the first user voice,
Judgment to determine whether or not the topic estimated by the estimation process is correct based on whether or not the second user voice spoken by the user in response to the output of the response voice shows negative content. Processing and executing,
A dialogue system characterized by that. The one or more processors
Refer to the similar keyword database that associates the correct keyword used in a specific topic with similar keywords similar to the correct keyword.
Further, a replacement process of replacing the similar keyword included in the voice recognition result of the first user voice with the correct answer keyword associated with the similar keyword in the similar keyword database is further executed.
The estimation process and the generation process are executed based on the voice recognition result after the replacement.
The dialogue system according to claim 1. The one or more processors
Refer to the similar keyword database associated with each of the plurality of topics including the specific topic.
In the estimation process, one of the plurality of topics is estimated.
2. The dialogue system according to claim 2. The one or more processors
If the topic determined to be correct by the determination process satisfies a predetermined condition, the transmission process of transmitting information about the user to the outside is further executed.
The dialogue system according to any one of claims 1 to 3, wherein the dialogue system is characterized in that. The one or more processors
Further executing the recording process of recording the first user voice in the memory,
In the transmission process, the access information for the first user voice recorded in the memory is included in the information about the user and transmitted.
The dialogue system according to claim 4, wherein the dialogue system is characterized in that. The one or more processors
In the estimation process, a topic related to health is estimated as the topic.
The dialogue system according to any one of claims 1 to 5, wherein the dialogue system is characterized in that. The dialogue robot included in the dialogue system according to any one of claims 1 to 6, comprising any one of the above-mentioned one or a plurality of processors.
The processor included in the dialogue robot is
At least the voice acquisition process for acquiring the first user voice and the second user voice and the voice output process for outputting the response voice are executed.
A dialogue robot characterized by that. A program for operating the dialogue system according to any one of claims 1 to 6, wherein the one or a plurality of processors execute each of the above processes. An estimation step in which one or more processors estimate a topic based on the first user voice spoken by the user.
A generation step in which the one or more processors generate a response voice in response to the first user voice.
The topic estimated by the estimation step is based on whether or not the second user voice spoken by the user in response to the output of the response voice by the one or more processors shows negative content. Including judgment steps to determine whether it is correct or not,
An information processing method characterized by that.

Images