JP2020016949A - Voice recognition data processing device, voice recognition data processing system, and voice recognition data processing method - Google Patents

Abstract

To improve consultation responding performance of a consultation counter by efficiently narrowing down necessary key words from consultation content of a client without omission at a consultation counter of an administrative agency, supporting simple creation of a summary of the consultation content, and lowering the load of the consultant.SOLUTION: A voice recognition data processing system determines respective contents of a plurality of items containing the genres of consultation contents according to comparison of the voice recognition results of voice data and keywords registered on the table, and narrows down the keywords among the plurality of keywords registered on the table for using the comparison according to the determined genre. The voice recognition data processing system generates a summary report of the consultation contents by using respective contents of the plurality of items determined on the basis of the keywords registered on the table or the narrowed down keywords.SELECTED DRAWING: Figure 5

Description

  The present disclosure relates to a voice recognition data processing device, a voice recognition data processing system, and a voice recognition data processing method.

  In a call at a call center that accepts inquiries about products and services, or at a consultation counter of a local government, etc., the uttered voice of the call is collected, and the uttered voice is recognized and converted into a character string (that is, text data). Processing voice recognition data processing systems have come to be used.

  As this type of speech recognition data processing system, for example, Japanese Patent Application Laid-Open No. H11-163873 discloses that a speech from a transmitting channel is speech-recognized by a first speech recognizing means, and a speech of a receiving channel transmitted through an exchange is transmitted to a second channel. 2. Description of the Related Art There is disclosed a call content transcription system that recognizes voice by two voice recognition means, converts each voice recognition result into a character string, and displays it. This call content transcription system is configured so that a displayed character string can be edited by a character string editing unit. As a result, the editor can transcribe (that is, convert to text) the contents of the call while confirming that the contents of the call on the transmitting and receiving sides are displayed in chronological order.

JP 2005-123869 A

  However, in Patent Literature 1, although it is considered to transcribe all of the call contents, it is not considered to create an outline (that is, a summary) of the call contents. Therefore, even if the configuration of Patent Literature 1 is applied to a case where a consultant responds to a consultation from a consultant at a consultation counter of an administrative agency (for example, a police station) or the like, the content of the consultation is converted into a text even if the consultant is applied. The output contents in which all the utterance contents of the counselor are converted to text are created. Therefore, when the counselor reviews the output contents later, it may be difficult to grasp the outline of the consultation content at a glance, and as a result, the convenience of the counselor may be reduced and the consultation counter may not improve the consultation response ability. There was sex.

  The present disclosure has been devised in view of the above-mentioned conventional circumstances, and efficiently narrows down necessary keywords from consultation contents of consultants without fail, and supports simple creation of summaries of consultation contents to assist consultants. It is an object of the present invention to provide a voice recognition data processing device, a voice recognition data processing system, and a voice recognition data processing method capable of reducing the load on the user and improving the consultation response capability of the consultation desk.

  The present disclosure relates to a voice recognition unit that recognizes voice data input from a microphone that collects a uttered voice of a consultant, a plurality of items including a type of consultation content from the consultant, and each of the items. And a memory that holds a table in which the keywords to be registered are associated with each other, and a plurality of items including the type of the consultation content according to a comparison between the voice recognition result of the voice data and the keywords registered in the table. A determination unit that determines each content and narrows down a keyword used for the comparison among a plurality of keywords registered in the table according to the determined type, and a keyword registered in the table or the narrowed down keyword. And generating a summary report of the consultation content using the content of each of the plurality of items determined based on the keyword that has been determined. Comprising a generating unit, and to provide a speech recognition data processing apparatus.

  Further, the present disclosure provides a voice recognition unit for recognizing voice data input from a microphone that collects a speech voice of a consultant, a plurality of items including a type of consultation content from the consultant, and each of the items. And a memory that holds a table in which a keyword corresponding to the keyword is registered, and a plurality of types including the type of the consultation content according to a comparison between the voice recognition result of the voice data and the keyword registered in the table. A determination unit that determines the content of each item and narrows down keywords used for the comparison among a plurality of keywords registered in the table, according to the determined type, and a keyword registered in the table or A summary report of the consultation content is generated using the content of each of the plurality of items determined based on the narrowed keywords. Comprising a generating unit that, the, to provide a speech recognition data processing system.

  Further, the present disclosure is a voice recognition data processing system having a memory that holds a table in which a plurality of items including a type of consultation content from a consultant and a keyword corresponding to each of the items are registered in association with each other. Voice recognition of voice data input from a microphone that collects the voice of the consultant, and, based on a comparison between a voice recognition result of the voice data and a keyword registered in the table, Determining the contents of each of the plurality of items including the type; narrowing down the keywords used for the comparison among the plurality of keywords registered in the table according to the determined type; registering in the table Of each of the plurality of items determined based on the selected keyword or the narrowed keyword Using, and a step of generating a summary report of the consulting detail, to provide a speech recognition data processing method.

  According to the present disclosure, it is possible to efficiently narrow down necessary keywords from the contents of consultation from a consultant without any omission, to support simple creation of a summary of the consultation contents, to reduce the load on a consultant, and to provide consultation at a consultation desk. The ability to respond can be improved.

Diagram showing the chart table maintained for speech recognition data processing Diagram showing an example of a summary report generated by speech recognition data processing FIG. 1 shows a configuration example of a speech recognition data processing system according to a first embodiment. FIG. 3 is a block diagram showing an example of the configuration of the speech recognition data processing system shown in FIG. 5 is a flowchart showing a main routine of voice recognition data processing according to the first embodiment. 5 is a flowchart showing a subroutine of the voice recognition process shown in FIG. 5 is a flowchart showing a subroutine of a keyword match determination process shown in FIG. 5 is a flowchart showing a subroutine of a chart table registration process shown in FIG. The figure which illustrates the chart table before the process of the chart table registration process shown in FIG. The figure which illustrates the chart table after the processing of the chart table registration processing shown in FIG. 5 is a flowchart showing a subroutine of a chart report generation process shown in FIG. FIG. 3 is a block diagram showing a modification of the speech recognition data processing system according to the first embodiment. 9 is a flowchart showing a main routine of voice recognition data processing according to the second embodiment. 13 is a flowchart showing a subroutine of a consultant presentation process shown in FIG. Flowchart showing counselor presentation processing according to Embodiment 3.

(Overview of Speech Recognition Data Processing Apparatus, Processing System and Processing Method)
Before giving a specific description of the speech recognition data processing system according to the present disclosure, an overview of the speech recognition data processing system will be described first with reference to FIGS. FIG. 1 is a diagram showing a chart table held for speech recognition data processing. FIG. 2 is a diagram illustrating an example of a summary report generated in the speech recognition data processing.

  In the first embodiment described in detail below, a plurality of items IT including a genre (type) of consultation content from a consultant and a plurality of keywords (including sentences) corresponding to each item IT and related to the item IT are also included. The table TB is registered and set in advance, and the table TB is stored in the memory 16 as a database (see FIGS. 1 and 4). The table TB further registers, for each keyword corresponding to the item IT, one or more keywords having low relevance as the genre of the consultation content. That is, the table TB holds keyword relevance information in which one or more keywords having low relevance in terms of genre are specified for each keyword. One or more keywords having low relevance for each keyword specified in the keyword relevance information are used when narrowing down keywords to be described later. Similarly, a chart table RT is held in the memory 16 as a format for inputting the contents of the consultation as text (see FIGS. 1, 4, 9, and 10). The format of the chart table RT is set in advance corresponding to each item IT registered in the table TB in advance.

  For example, when the consultation desk is a police station among the administrative organizations, the contents of the consultation are mainly assumed to relate to, for example, crimes and annoying acts. In the chart RT, as shown in FIG. 1, the item IT of the consultation content includes “genre (type)”, “consultant”, “partner”, “damage status”, and “ "Request", "Contents of Police Proposal", "Agreement / Understanding of Counselor", "Next Action", etc. are set. Further, a content input column CT for inputting specific consultation content for each item IT is also provided in the chart table RT. Further, in the chart table RT, a keyword input field K1 for inputting a keyword obtained based on the voice data of the consultant C1 and a keyword obtained based on the voice data of the consultant C2 are input. Is provided for each item IT (see FIGS. 9 and 10). A plurality of keywords respectively entered in the consultant-side keyword input field K1 and the consultant-side keyword input field K2 are selected from keywords specified in the table TB and registered in advance, as described later. The information may be appropriately updated (added, changed, or deleted) according to the frequency of use by a learning process using artificial intelligence (AI).

  Note that, in the initial state of the chart RT, the content input column CT of each item IT is set to at least a blank column. Further, the chart table RT is merely an example, and is not limited to the contents shown in FIG. 1, and may be appropriately changed and set according to the nature of the consultation desk. The same applies to the following drawings, which are examples for promoting understanding of the present disclosure.

  The voice recognition data processing system 1 according to the first embodiment performs voice recognition of the uttered voices of the consultant C1 and the consultant C2 for each phrase, and compares the voice recognition result with the keyword registered in the table TB. Then, the contents of each of the above-described items IT including the genre of the consultation contents are determined, and texts are input to the respective content columns CN of the chart table RT.

  As a specific example, in FIG. 1, the keyword of "neighborhood trouble" is included in the content of the "genre" having the first item number NO, and the content is first noted in the content of the sixth "content of proposal from police". We will be careful from the beginning. We will input the keyword "We will be careful." In the 8th "Next action," the keyword "Tell me the day to be careful" will be input as text as a result of speech recognition You. At this time, the speech recognition data processing system 1 uses the table TB to narrow down keywords used for comparison among a plurality of keywords registered in the table TB according to the determined genre. By this narrowing down, it is possible to speed up the keyword matching judgment on the speech recognition result.

  Each item IT of the medical chart table RT is sequentially input for each phrase of the voice data, and each content is sequentially determined as a list indicating the entire consultation content. In the process, each content column CN is filled, and the chart table RT is completed. At this time, as shown in FIG. 2, a summary report of the consultation contents is generated based on the contents of each of the determined items IT.

  As a specific example, as shown in FIG. 2, based on the chart RT shown in FIG. 1, a summary report AB titled "Title: Neighborhood trouble" and "Consultant: OO" is generated. The content of the summary report AB is "I'm in trouble with a neighborhood trouble .... (omitted) ... As a proposal from the police, why don't you start with the place to be careful first. Propose that the police pay attention (with underline). .. (omitted) ... As the next action, contact the day on which the police should pay attention (underlined). " In the contents of the summary report AB, the contents corresponding to the items requiring confirmation are underlined as described later. As described above, the summary report AB may be generated in a sentence format as shown in FIG. 2, or may be generated in a table format such as the chart table RT shown in FIG.

  As described above, in the first embodiment, the creation of the summary report AB allows the consultation counter of the administrative agency or the like (the police station in this description) to efficiently transmit the necessary keywords from the consultation contents from the consultant C1 at the consultation counter. It is possible to support narrowing down and simple creation of a summary of consultation contents. Therefore, in order to support the creation, the speech recognition data processing device 10, the speech recognition data processing system 1, and the speech recognition data processing method have specific configurations described in a plurality of embodiments described below.

  Hereinafter, an embodiment that specifically discloses a speech recognition data processing device, a speech recognition data processing system, and a speech recognition data processing method according to the present disclosure will be described in detail with reference to the drawings as appropriate. However, an unnecessary detailed description may be omitted. For example, a detailed description of a well-known item or a redundant description of substantially the same configuration may be omitted. This is to prevent the following description from being unnecessarily redundant and to facilitate understanding by those skilled in the art. The accompanying drawings and the following description are provided to enable those skilled in the art to fully understand the present disclosure, and are not intended to limit the claimed subject matter.

  The “unit” or “apparatus” in the first, second, and third embodiments is not limited to a physical configuration realized simply by hardware, and the functions of the configuration are realized by software such as a program. Including things. Further, the functions of one configuration may be realized by two or more physical configurations, or the functions of two or more configurations may be realized by, for example, one physical configuration.

(Embodiment 1)
The configurations of the speech recognition data processing device 10 and the speech recognition data processing system 1 according to the first embodiment will be described with reference to FIGS. FIG. 3 is a schematic diagram illustrating an example of a configuration of the speech recognition data processing system 1 according to the first embodiment. FIG. 4 is a block diagram showing an example of the configuration of the speech recognition data processing system 1 shown in FIG.

  As illustrated in FIGS. 3 and 4, the voice recognition data processing device 10 and the voice recognition data processing system 1 according to the first embodiment are exemplified for use at a consultation counter installed in an administrative agency (for example, a police station). Is done. For example, it is assumed that a consultant C1 who is a local resident goes to a consultation counter of a police station and meets a consultant C2 such as a police officer to directly consult a crime or annoyance.

  In the first embodiment, a plurality of microphones M1 and M2 and a voice recognition data processing device 10 are installed at the consultation desk, and voices of the consultant C1 and the consultant C2 are collected through the microphones M1 and M2. Then, the voice data is input to the voice recognition data processing device 10 to perform voice recognition. In the first embodiment, description will be made assuming use at a consultation counter of a police station, but the present invention is not limited to this. In addition, it is applicable to a consultation desk at a city hall, a consultation desk at a medical institution such as a hospital, and a consultation desk at a private institution without being limited to a public institution.

  In the voice recognition data processing system 1 according to the first embodiment, a plurality of (for example, two) microphones M1 and M2 for collecting speech voices of the consultant C1 and the consultant C2, and input from the microphones M1 and M2 are provided. A speech recognition data processor 10 for generating a summary report AB (see FIG. 2) of consultation contents based on the speech data, and a speech recognition server 20 for automatically recognizing speech based on speech data transmitted from the speech recognition data processor 10. , And a network NW that connects the voice recognition data processing device 10 and the voice recognition server 20.

  The plurality of microphones M1 and M2 collect voices spoken by the consultant C1 and the consultant C2, and output the voices as electrical signal voice data. The voice data is output to the voice recognition data processing device 10 as electrical vibration waveform data. In the first embodiment, two microphones M1 and M2 are installed, but the number of microphones is not limited to two and may be one. In this case, one microphone M1 may be configured to be able to collect the uttered voices of both the consultant C1 and the consultant C2 so as to be identifiable.

  The voice recognition data processing device 10 is configured by a general-purpose personal computer, and includes a processor 11, a memory 16, a display 17, an input / output interface circuit 18, and a communication circuit 19. In FIG. 4, the input / output interface circuit is indicated as “input / output I / F” for convenience. Note that the voice recognition data processing device 10 is not limited to a personal computer, and various devices such as a tablet terminal and a smartphone may be used as long as they have the above-described various circuits.

  The memory 16 includes a semiconductor memory such as a RAM (Random Access Memory) and a ROM (Read Only Memory). The memory 16 stores and holds a predetermined program executed by the processor 11, various setting parameters, and the like, and temporarily stores digitally converted audio data and data generated intermediately when the processor executes the various programs. To memorize.

  Further, in the first embodiment, as described above, the memory 16 stores the plurality of items IT including the genre of the consultation content from the consultant C1 and the plurality of keywords corresponding to each item IT and related to the item IT. Stores and holds a table TB registered in advance in association with the table TB. That is, as described above, in the table TB, a plurality of keywords are set in advance for each item IT of the consultation content, and the plurality of keywords are group-registered and stored for each item IT. The memory 16 also stores a chart table RT (see FIG. 9) as a format for inputting consultation contents. The table TB and the chart table RT may be stored and held in a storage circuit such as a storage device such as a solid state drive (SSD) or a hard disk drive (HDD) instead of a semiconductor memory. Further, as described above, the table TB further registers, for each keyword corresponding to the item IT, another keyword having low relevance as the genre of the consultation content.

  The display 17 includes a liquid crystal display, an organic EL (Electro Luminescence) display, and the like. The display 17 switches the sound data, the chart table RT, the summary report AB, and the like as appropriate on the display unit and displays the data to at least the consultant C2.

  The input / output interface circuit 18 is an audio interface that converts audio data acquired by the microphones M1 and M2 into digital signals that can be subjected to various signal processing. At the time of the digital conversion, the input / output interface circuit 18 converts the analog audio data into a digital signal using a predetermined quantization bit and a sampling frequency. The sampling frequency is set to, for example, 48 kHz. Also, the input / output interface circuit 18 includes a buffer circuit, and this buffer circuit buffers audio data for a predetermined time in order to convert the audio data of the microphones M1 and M2 into digital data. The input / output interface circuit 18 transmits the digitally converted audio data to the processor 11 and the display 17.

  The communication circuit 19 has a wireless communication function, is connected to the network NW via a predetermined wireless communication line, and performs wireless communication with the voice recognition server 20. That is, the voice recognition data processing device 10 is connected to the voice recognition server 20 via the communication circuit 19 and the network NW, and is configured to be able to transmit and receive predetermined data such as voice data. The communication circuit 19 can use a mobile phone network (cellular network), a wireless LAN, or the like as a wireless communication circuit. The communication circuit 19 is not limited to wireless communication, and may be configured to enable wired communication using a wired LAN or the like.

  As the processor 11, for example, a CPU (Central Processing Unit), an MPU (Micro Processor Unit), a DSP (Digital Signal Processor), or an FPGA (Field Programmable Gate Array) is used. The processor 11 reads out a program and setting parameters stored in the memory 16 as appropriate, and executes a predetermined process according to the read out program.

  As a functional configuration, the processor 11 compares the speech recognition result with a keyword registered in the table TB with a speech recognition unit 12 that recognizes speech data obtained from the microphones M1 and M2, and compares each of the items IT. A keyword match determination unit 13 (an example of a determination unit) that determines the contents, and a chart report generation unit 14 (an example of a generation unit) that generates a summary report AB of the consultation content based on the determination of each item IT are registered and registered. And an AI (Artificial Intelligence) processing unit 15 that learns a plurality of keywords and the determination result of the keyword match determination unit 13 and appropriately updates the registered contents of the keywords in the table TB.

  The voice recognition unit 12 acquires digitally converted voice data from the input / output interface circuit 18. The voice recognition unit 12 performs voice recognition on the obtained voice data based on, for example, a predetermined corpus, and converts the voice data into text data (character string). The conversion and interpretation of the text data are executed according to a predetermined automatic speech recognition algorithm. The automatic speech recognition algorithm is generated using, for example, an artificial intelligence (AI) engine implemented in the AI processing unit 15.

  Here, the speech recognition unit 12 of the speech recognition data processing device 10 is limited by the processing capability (that is, performance) of the processor of the speech recognition data processing device 10. The performance and accuracy of the voice recognition unit 12 may be limited because the voice recognition unit 12 follows the automatic voice recognition algorithm generated by the AI processing unit 15 of the processor 11. Therefore, in the first embodiment, the voice recognition server 20 capable of acquiring voice data of the contents of consultation at various consultation desks is normally used via the network NW, and a voice recognition process described later (for example, FIG. (See step S20 shown). However, when the connection of the network NW is not good, for example, when the traffic volume of the network NW becomes excessive, the voice recognition unit 12 of the voice recognition data processing device 10 functions instead of the voice recognition server 20 to perform voice recognition on the voice data. May do it. Similarly, the speech recognition server 20 generates an automatic speech recognition algorithm by learning artificial intelligence and performs speech recognition.

  The learning of the artificial intelligence executed by the speech recognition data processing device 10 and the speech recognition server 20 may be executed using one or more statistical classification techniques. Statistical classification techniques include linear classifiers, support vector machines, quadratic classifiers, kernel density estimation, decision trees, artificial neural networks. Networks (artificial neural networks), Bayesian techniques and / or networks, hidden Markov models, binary classifiers, multi-class classifiers, Examples include a clustering technique, a random forest technique, a logistic regression technique, a linear regression technique, and a gradient boosting technique. However, the statistical classification technique used is not limited to these.

  The voice recognition server 20 or the voice recognition unit 12 of the voice recognition data processing device 10 sequentially transmits the text data as the voice recognition result to the keyword match determination unit 13. The keyword match determination unit 13 acquires the speech recognition result for each phrase, and determines whether or not there is a match with a keyword registered in advance in the table TB. That is, the keyword match determination unit 13 performs comparison with a plurality of keywords registered in the table TB every time the voice recognition result of voice data corresponding to one phrase of the uttered voice of the consultant C1 and the consultant C2 is obtained. Do. Further, the keyword match determination unit 13 uses the table TB to narrow down keywords to be used for comparison among a plurality of keywords registered in the table TB according to the determined genre.

  The medical chart report generation unit 14 generates a consultation content summary report AB using the contents of the plurality of items IT determined based on the keywords registered in the table TB or the narrowed keywords (see FIG. 2). ).

  Next, the procedure of the voice recognition data processing method according to the first embodiment will be described with reference to FIGS. FIG. 5 is a flowchart showing a main routine of the voice recognition data processing according to the first embodiment. FIG. 6 is a flowchart showing a subroutine of the voice recognition processing S20 shown in FIG. FIG. 7 is a flowchart showing a subroutine of the keyword match determination processing S30 shown in FIG. FIG. 8 is a flowchart showing a subroutine of the chart table registration processing S40 shown in FIG. FIG. 9 is a diagram exemplifying the chart table RT before the processing of the chart table registration processing S40 shown in FIG. FIG. 10 is a diagram illustrating the medical chart table RT after the processing of the medical chart table registration processing S40 illustrated in FIG. 5. FIG. 11 is a flowchart showing a subroutine of the chart report generation processing S50 shown in FIG.

  When responding to the consultant C1, the consultant C2 operates the voice recognition data processing system 1 (see FIG. 3). Each of the microphones M1 and M2 starts collecting speech voices of the consultant C1 or the consultant C2, and outputs analog voice data to the voice recognition data processing device 10. As shown in FIG. 5, the voice recognition data processing device 10 acquires the input voice data for each phrase by the input / output interface circuit 18, and displays the vibration waveform of the voice data on the display 17 (S10). ).

  Next, as shown in FIGS. 5 and 6, a subroutine of the voice recognition process is started (S20). In the voice recognition process S20, the input / output interface circuit 18 buffers the voice data in the buffer circuit, converts the voice data into digital data, and outputs the digital data to the processor 11 (S21). The processor 11 transmits the digitally converted voice data to the voice recognition server 20 through the communication circuit 19 (S22). The voice recognition server 20 receives the voice data, recognizes the received voice data according to an automatic voice recognition algorithm, and converts the voice data into text data as a voice recognition result (S23). At this time, the voice recognition server 20 transmits the voice recognition result to the voice recognition data processing device 10 for each phrase of the uttered voice. The voice recognition data processing device 10 receives the voice recognition result of the voice recognition server 20 through the communication circuit 19 (S24). With this reception, the subroutine of the voice recognition processing ends.

  Next, as shown in FIG. 5 and FIG. 7, after the voice recognition processing S20 is completed, a subroutine of the keyword matching determination processing is started (S30). In the keyword match determination process S30, the keyword match determination unit 13 of the processor 11 acquires a voice recognition result of voice data corresponding to one phrase of the uttered voice of the consultant C1 and the consultant C2 (S31). At the time of the acquisition, the keyword match determination unit 13 decomposes the speech recognition result for each word (S32). The word here is not limited to a word, but may be a sentence such as a phrase or a section. The keyword match determination unit 13 compares the speech recognition result with a plurality of keywords registered in the table TB in advance and determines whether or not they match, based on the speech recognition corresponding to the utterance contents of each of the consultant C1 and the consultant C2. The results are compared individually (S33). As a result of the comparison, the keyword match determination unit 13 determines whether or not the keyword matches the keyword registered in the table TB (S34). If it is determined that they do not match, the subroutine of the keyword match determination processing S30 ends. Note that the keyword matching determination unit 13 performs supplementary conversion as appropriate, for example, to convert the uttered voice of "we" to "police" or the uttered voice of "me" to "consultant" at the time of keyword match determination (S34). First person conversion according to the situation may be configured.

  Further, by the termination processing of the keyword match determination processing S30, the voice recognition data processing system 1 allows the consultant C1 or the consultant C2 to have no relation to the consultation content (for example, chat or small talk) while the microphones M1 and M2 are collecting sound. Etc.), and proceed to the next chart table registration process (step S40). This makes it possible to suppress the content unrelated to the consultation content, that is, noise information from being mixed into the chart table RT and the summary report AB.

  On the other hand, if it is determined that they match, the keyword match determination unit 13 further determines whether or not the matched voice recognition result matches a keyword corresponding to the genre of the consultation content (S35).

  If it is determined that they do not match, the subroutine of the keyword match determination processing S30 ends. On the other hand, when it is determined that the keyword matches the keyword corresponding to the genre, the keyword match determination unit 13 determines whether the genre of the consultation content of the consultant C1 is, for example, “neighboring trouble”, “stalker damage”, or “fraud”. It is determined whether it is "damage" or the like, and the result of the determination is input to the chart RT. Further, the keyword match determination unit 13 narrows down the keywords used in the above-described step S33 among the plurality of keywords registered in the table TB according to the determined genre (S36). For example, as described above, the table TB contains, for each keyword corresponding to the item IT, keyword relevance information in which one or more keywords having low relevance from the viewpoint of the genre of the consultation content are specified. keeping. For example, it is considered that the keyword of the genre “Stalker” has low relevance to the keyword “XX circle transfer”. Also, the keyword “genre fraud” is considered to have low relevance to the keyword “cut”. As described above, one or more keywords having low relevance registered for each keyword are used when the keyword matching determination unit 13 narrows down the keywords. Thus, the subroutine of the keyword match determination processing S30 is completed.

  Next, as shown in FIGS. 5 and 8, after the keyword match determination processing S30 is completed, a subroutine of the chart table registration processing is started (S40). In the chart table registration process S40, the keyword match determination unit 13 of the processor 11 refers to the table TB in the memory 16 and searches for an item IT corresponding to the keyword determined to match in step S33 (S41). As a result of the search, the keyword match determination unit 13 determines whether a keyword has been input in the corresponding item IT (S42). If it is determined that the keyword has not been input, the keyword matching determination unit 13 determines and inputs the matching keyword into the content column CN of the corresponding item IT (S45).

  Specifically, as shown in FIG. 9, in the words of the speech recognition result based on the utterance of the consultant C1, the keyword “the person in the next room” in the consultant-side keyword input field K1 matches in step S33. When it is determined that the user has performed the search, the item (genre) “neighborhood trouble” corresponding to the keyword “the person in the next room” registered in the table TB is input to the genre content column CN.

  Further, as another example, when it is determined in step S33 that the keyword “Stalker” in the keyword input field K2 of the consultant matches in the words of the voice recognition result based on the utterance of the consultant C2, the table TB is displayed. An item (genre) “stalker” corresponding to the registered keyword “stalker” is input in the genre content column CN. Further, the chart table RT being input by the voice recognition data processing system 1 is displayed on the display 17 of the voice recognition data processing device 10, and the consultant C2 can check the input status during the consultation. .

  On the other hand, when it is determined that the corresponding item IT has been input (that is, when a plurality of contents exist in any one of the plurality of items IT), the keyword match determination unit 13 is stored in the table TB. By referring to the keyword relevance information, it is determined whether or not a plurality of inputted contents are inconsistent in relation to the genre of the corresponding item IT (S43). If it is determined that there is no contradiction, the subroutine of the chart table registration processing S40 ends.

  On the other hand, if it is determined that there is a contradiction, the keyword match determination unit 13 causes the memory 16 to store that fact as a confirmation required item (S44). The keyword match determination unit 13 determines and inputs a keyword for the corresponding item IT for each sequentially acquired voice recognition result (S45). Further, based on the input keyword, a heading for the content of each item IT is automatically recognized, and the heading is input to the content column CN of the content of the chart table RT (see FIG. 10). Thus, the subroutine of the chart table registration process S40 ends.

  Next, as shown in FIGS. 5 and 11, after the chart table registration processing S40 is completed, a subroutine of the chart report generation processing is started (S50). In the medical chart report generation processing S50, the medical chart report generation unit 14 of the processor 11 refers to the medical chart table RT stored in the memory 16 and searches for the item IT whose contents have been input (S51). As a result of the search, the medical record report generator 14 generates a text corresponding to the input item IT (S52). At this time, the chart report generation unit 14 also refers to the required confirmation items stored in the memory 16 and, when it is determined that any of the items IT is inconsistent, requires the contents of the corresponding item IT. Generate text so that it can be identified as a confirmation item. Specifically, the medical chart report generation unit 14 adds an underline to the content corresponding to the item requiring confirmation to make it identifiable (see FIG. 2).

  Further, the chart report generation unit 14 corrects the generated text so that the sense of incongruity does not occur in the meaning before and after (S53), and supposes and inserts the subject in the text in which the subject is missing (S54). At this time, when the subject is guessed, the chart report generation unit 14 similarly generates a text that can be identified by underlining the item to be checked.

  As described above, the chart report generating unit 14 generates the summary report of the consultation content AB using the contents of the plurality of items IT determined based on the keyword (see FIG. 2). The processor 11 displays the summary report AB on the display 17 via the input / output interface circuit 18 (S56).

  As shown in FIG. 5, after the chart report generation processing S50 ends, it is determined whether an end trigger for the main processing has been input (S60). As a result of the determination, unless a termination trigger is input, a series of processing steps from step S10 to step S50 are repeatedly executed. When the end trigger is input, for example, when the consultation at the consultation desk is completed and the switch of the voice recognition data processing system is turned off (OFF) by the consultant C2, the main processing of the voice recognition data processing system 1 ends. .

  As described above, the voice recognition data processing device 10 and the voice recognition data processing system 1 according to the first embodiment provide the voice recognition unit 12 that recognizes voice data input from the microphone M1 that collects the voice of the consultant C1. Is provided. In the voice recognition data processing device 10 and the voice recognition data processing system 1, a plurality of items IT including the genre (type) of the consultation content from the consultant C1 and keywords corresponding to the respective items IT are registered in association with each other. A memory 16 for holding the table TB is provided. The voice recognition data processing device 10 and the voice recognition data processing system 1 perform the respective content of the plurality of items IT including the genre of the consultation content in accordance with the comparison between the voice recognition result of the voice data and the keyword registered in the table TB. And a keyword match determination unit 13 (an example of a determination unit) that narrows down keywords used for comparison among a plurality of keywords registered in the table TB according to the determined genre. The voice recognition data processing device 10 and the voice recognition data processing system 1 use the content of each of the plurality of items IT determined based on the keywords registered in the table TB or the narrowed-down keywords to generate a summary report of the consultation content. A medical chart report generation unit 14 (an example of a generation unit) that generates an AB is provided.

  Therefore, in the first embodiment, the voice recognition data processing device 10 determines the content of each of the plurality of items including the type of the consultation content according to the comparison between the voice recognition result of the voice data and the keyword registered in the table. decide. The speech recognition data processing device 10 narrows down keywords to be used for comparison among a plurality of keywords registered in the table TB according to the determined type, and determines based on the keywords registered in the table TB or the narrowed down keywords. A summary report AB of the consultation content is generated using the contents of each of the plurality of items. Thus, the speech recognition data processing device 10 can efficiently narrow down the necessary keywords from the consultation contents from the consultant at the consultation counter of the administrative organization or the like by creating the summary report AB, and summarize the consultation contents. Support simple creation and reduce the load on consultants. As a result, the consultation service can improve the consultation response ability.

  The speech recognition data processing device 10 and the speech recognition data processing system 1 further include a processor 11 (an example of a control unit) that displays the generated summary report AB of the consultation content on the display 17. Thereby, the consultant C2 can check the contents of the consultation while receiving the consultation at the consultation counter. Therefore, the voice recognition data processing device 10 further reduces the load on the consultant C2 and also consults the consultant C2 regarding the contents of the consultation. The quality of the question can also be improved.

  Further, in the speech recognition data processing device 10 and the speech recognition data processing system 1, the keyword match determination unit 13 (an example of a determination unit) determines whether or not a plurality of contents exist in any one of the plurality of items IT. It is determined whether or not a plurality of contents are inconsistent in relation to the corresponding item IT. When it is determined that the plurality of contents are inconsistent in relation to the corresponding item IT, the chart report generation unit 14 (an example of a generation unit) can identify at least one of the plurality of contents as the item requiring confirmation. Generate a summary report AB. Thereby, the speech recognition data processing apparatus 10 improves the consistency and consistency of the summary report AB, improves the quality of the summary report, and makes it easy for the consultant C2 to grasp the consultation content at a glance when reviewing it later. Can be

  In the speech recognition data processing device 10 and the speech recognition data processing system 1, the chart report generation unit 14 (an example of a generation unit) generates a summary report AB by adding an underline to the content corresponding to the item requiring confirmation. Accordingly, the voice recognition data processing device 10 enhances the visibility of the items requiring confirmation in the summary report AB, so that the consultant C2 during consultation is urged to reconfirm, and the summary report is easily modified after consultation. Or you can.

  In the voice recognition data processing device 10 and the voice recognition data processing system 1, the keyword match determination unit 13 (an example of a determination unit) acquires a voice recognition result of voice data corresponding to one phrase of the uttered voice of the consultant C1. Each time the keyword is compared with the keyword registered in the table TB. As a result, during the consultation at the consultation desk, the voice recognition and the keyword comparison are performed and are sequentially input and displayed on the chart table RT, so that the consultant C2 grasps the input status of the chart table RT and recognizes the consultant. C1 can be handled. For this reason, the consultation response ability at the consultation desk can be further enhanced.

(Modification of First Embodiment)
As a modified example of the first embodiment, the voice recognition data processing system 1 may further include an AI processing server 30 that complements or substitutes for the function of voice recognition data processing. The AI processing server 30 is connected to the network NW and provided with enhanced artificial intelligence performance. The AI processing server 30 is configured by a general-purpose server system (high-performance computer), and includes a processor 31, a memory 36, and a communication circuit 39, similarly to the speech recognition data processing device 10. You.

  Similarly, the processor 31 of the AI processing server 30 has a keyword match determination unit 33, a chart report generation unit 34, and an AI processing unit 35 as functional configurations. However, these AI processing servers 30 have enhanced artificial intelligence performance, and are configured to be able to acquire voice data at a plurality of consultation desks. Therefore, the model of the artificial intelligence can be constructed with higher accuracy than the speech recognition data processing device 10. Therefore, in the modification of the first embodiment, it is possible to reduce the load on the consultant C2 by supporting the creation of the summary of the consultation content faster and more accurately.

(Embodiment 2)
Next, a procedure of a voice recognition data processing method of the voice recognition data processing device 10 and the voice recognition data processing system 1 according to the second embodiment will be described with reference to FIGS. FIG. 13 is a flowchart showing a main routine of voice recognition data processing according to the second embodiment. FIG. 14 is a flowchart showing a subroutine of the consultant presenting process S70 shown in FIG. The same or equivalent parts as those in the first embodiment will be given the same or equivalent reference numerals in the drawings and the like, and the description thereof will be omitted or simplified.

  As shown in FIGS. 13 and 14, in the second embodiment, after the chart report generation processing (S50) ends, a subroutine of the consultant presentation processing is started (S70). After the end of the consultant presenting process S70, the processor 11 determines whether or not an end trigger has been input as in the first embodiment (S60). In addition, the counselor presentation processing S60 is mainly executed by the AI processing unit 15 (an example of the presentation processing unit) of the processor 11, but is not limited thereto. The AI processing unit 35 of the AI processing server 30 performs the processing. May be configured to be able to assist or substitute.

  In the consultant presenting process S70, the AI processing unit 15 of the processor 11 searches for the presence or absence of the item IT of the item IT whose content has not yet been determined according to a predetermined priority order according to the determination status of the chart table RT. (S71). As a result of the search, the AI processing unit 15 determines whether there is an undetermined item IT (S72). When it is determined that there is no undetermined item IT, the consultant presenting process S70 ends.

  On the other hand, when it is determined that there is an undetermined item IT, the AI processing unit 15 presents information about the undetermined item IT through the display 17 as “item to be asked next by the consultant (C2)”. For example, when the genre of the consultation content has been determined and the item IT of the damage status has not been determined, the AI processing unit 15 presents “Please ask for the damage status” to the consultant C2. After the presentation, the consultant presentation process S70 ends. The information on the priority is set in advance, and is stored in the memory 16 of the speech recognition data processing device 10.

  As described above, the voice recognition data processing device 10 and the voice recognition data processing system 1 according to the second embodiment determine whether or not there is an item IT whose content has not yet been determined according to the determination status of each of the plurality of items IT. And an AI processing unit (an example of a presentation processing unit) for displaying information on the undetermined item IT on the display 17. Thereby, the consultant C2 can hear the contents of the consultation from the consultant C1 without omission, so that the quality of the consultation contents can be further improved, and the consultation service ability of the consultation counter can be further improved. Other configurations and operational effects are the same as those in the first embodiment.

(Embodiment 3)
Next, a procedure of a speech recognition data processing method of the speech recognition data processing device 10 and the speech recognition data processing system 1 according to the third embodiment will be described with reference to FIG. FIG. 15 is a flowchart showing a consultant presentation process S70 according to the third embodiment. The same or equivalent parts as those in the first and second embodiments are given the same or equivalent reference numerals in the drawings and the like, and the description thereof is omitted or simplified.

  As shown in FIG. 15, in the counselor presentation process S70 of the third embodiment, the AI processing unit 15 (an example of a presentation processing unit) of the voice recognition data processing device 10 proposes from the counselor C2 to the counselor C1. A search is made to see if the necessary information is available (S75). As a result of the search, the AI processing unit 15 determines whether or not the number of determined contents of each of the plurality of items IT including the genre of the consultation content is a predetermined value, and at least whether the content of the genre is determined (S76). . For example, the AI processing unit 15 determines whether the first to fifth contents of the item number NO of the item IT illustrated in FIG. 1 have been determined. If it is determined that no decision has been made, the consultant presenting process S70 ends.

  On the other hand, if it is determined that the decision has been made, the AI processing unit 15 generates the proposal content of the measure corresponding to the genre based on the decided content (S77). At this time, the AI processing unit 15 uniquely determines the content of the proposal in accordance with a predetermined hash table in which rules (regulations) are prescribed in advance. The AI processing unit 15 presents the generated text of the proposal content on the display 17 (S78), and the consultant presenting process S70 ends.

  Note that the AI processing unit 15 is not limited to a method according to a rule base such as a hash table, but uses artificial intelligence to refer to a past history and machine-learns the proposal contents, and obtains the proposed contents by the machine learning. The determination may be made according to a learning model. The learning model may be generated by the AI processing server 30. In this case, since a learning model is generated from the consultation contents of a plurality of consultation desks, a highly accurate learning model can be generated.

  As described above, in the voice recognition data processing device 10 and the voice recognition data processing system 1 according to the third embodiment, the number of determined contents of each of the plurality of items IT including the genre (type) of the consultation content is the predetermined value, and An AI processing unit 15 (an example of a presentation processing unit) that generates proposal contents of measures corresponding to the contents of the genre and displays it on the display 17 when at least the contents of the genre is determined. Thereby, since the proposal content is appropriately and automatically derived based on the consultation content and is presented to the consultant C2, the support for the consultant C1 is facilitated and the load on the consultant C2 can be further reduced. Other configurations and operational effects are the same as those in the first and second embodiments.

  Although various embodiments have been described with reference to the drawings, it is needless to say that the present disclosure is not limited to such examples. It is clear that those skilled in the art can conceive various changes, modifications, substitutions, additions, deletions, and equivalents within the scope of the claims. Naturally, it is understood that they belong to the technical scope of the present disclosure. Further, the components in the above-described various embodiments may be arbitrarily combined without departing from the spirit of the invention.

  This disclosure reduces the burden on counselors by efficiently narrowing down the required keywords from the contents of consultations from consultants at consultation counters of administrative agencies, etc., and supporting simple creation of summaries of consultation contents. Therefore, the present invention is useful as a voice recognition data processing device, a voice recognition data processing system, and a voice recognition data processing method capable of improving consultation ability.

1 Speech Recognition Data Processing System 10 Speech Recognition Data Processing Device 11 Processor 12 Speech Recognition Unit 13 Keyword Match Determination Unit 14 Medical Record Report Generation Unit 15 AI Processing Unit 16 Memory 17 Display 18 Input / Output Interface Circuit 19 Communication Circuit 20 Speech Recognition Server 30 AI Processing server 31 Processor 33 Keyword match determination unit 34 Medical record report generation unit 35 AI processing unit 36 Memory 39 Communication circuit M1 Microphone M2 Microphone NW Network TB Table AB Summary report

Claims (9)

A voice recognition unit that recognizes voice data input from a microphone that collects speech of a consultant,
A memory that holds a table in which a plurality of items including the type of consultation content from the consultant and a keyword corresponding to each of the items are registered in association with each other;
According to the comparison between the voice recognition result of the voice data and the keyword registered in the table, the content of each of the plurality of items including the type of the consultation content is determined, and according to the determined type, A determination unit that narrows down keywords used for the comparison among a plurality of keywords registered in the table;
A generating unit that generates a summary report of the consultation content, using the content of each of the plurality of items determined based on the keyword registered in the table or the narrowed keyword,
Voice recognition data processing device. A control unit for displaying the generated summary report of the consultation content on a display, further comprising:
The speech recognition data processing device according to claim 1. The determination unit, when there is a plurality of contents in any of the plurality of items, determines whether the plurality of contents are inconsistent in relation to a corresponding item,
The generating unit, when it is determined that the plurality of contents are inconsistent with a corresponding item, the summary report so that at least one of the plurality of contents can be identified as a required item. Generate,
The speech recognition data processing device according to claim 1. The generation unit generates the summary report by adding an underline to the content corresponding to the item requiring confirmation,
The voice recognition data processing device according to claim 3. The determination unit performs a comparison with a keyword registered in the table each time a voice recognition result of the voice data corresponding to one phrase of the utterance voice of the consultant is obtained.
The speech recognition data processing device according to claim 1. A presentation processing unit that searches for the presence or absence of an item whose content has not yet been determined, according to the determination status of the content of each of the plurality of items, and displays information about the undetermined item on a display.
The speech recognition data processing device according to claim 1. When the number of determined contents of each of the plurality of items including the type of the consultation content is a predetermined value, and at least the content of the type is determined, a proposal content of a measure corresponding to the content of the type is generated. A presentation processing unit for displaying on a display,
The speech recognition data processing device according to claim 1. A voice recognition unit that recognizes voice data input from a microphone that collects speech of a consultant,
A memory that holds a table in which a plurality of items including the type of consultation content from the consultant and a keyword corresponding to each of the items are registered in association with each other;
According to the comparison between the voice recognition result of the voice data and the keyword registered in the table, the content of each of the plurality of items including the type of the consultation content is determined, and according to the determined type, A determination unit that narrows down keywords used for the comparison among a plurality of keywords registered in the table;
A generating unit that generates a summary report of the consultation content, using the content of each of the plurality of items determined based on the keyword registered in the table or the narrowed keyword,
Voice recognition data processing system. In a voice recognition data processing system having a memory that holds a table in which a plurality of items including a type of consultation content from a consultant and a keyword corresponding to each of the items are registered in association with each other,
Voice recognition of voice data input from a microphone that collects the utterance voice of the consultant;
Determining the content of each of a plurality of items including the type of the consultation content according to a comparison between the voice recognition result of the voice data and the keyword registered in the table;
Narrowing down a keyword used for the comparison among a plurality of keywords registered in the table according to the determined type;
Generating a summary report of the consultation content by using the content of each of the plurality of items determined based on the keyword registered in the table or the narrowed keyword,
Voice recognition data processing method.

Images