JP2010094243A - Automatic inquiry system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an automatic inquiry system for guiding a patient to give assumed answers when he or she gives unexpected answers. <P>SOLUTION: The automatic inquiry system includes: a microphone; a speaker; a voice recognition means; a voice output means; a question information storage means wherein the question information of a question sentence to a patient in an inquiry interaction scenario is stored; an inquiry result storage means for storing the answer information of a selected answer to the question sentence in relation to the question sentence; a question read means for successively reading, sounding and outputting the question information; an answer discrimination means for discriminating whether the answer information of the patient is the assumed answer or the unexpected answer; a guide information storage means for storing guide information of a guide sentence for guiding the patient to correctly select an answer to the question sentence corresponding to the type of choices of each question sentence; an answer preservation means for preserving the answer information in the inquiry result storage means in the case of the assumed answer; and a guide output means for sounding and outputting the guide information corresponding to the type of the choices of the question sentence in the case of the unexpected answer. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to an automatic inquiry system that automatically performs an inquiry of a patient in a hospital or the like.

  As a conventional automatic inquiry system, the thing of patent documents 1-3 is publicly known.

  In Patent Document 1, a server that stores first data related to a question set in advance by a health care provider and second data related to the first data, and a server that stores the second data related to the first data from the server via the communication network. 1 data is read out and displayed on the first display means, and a reply made by the voice or image to the question displayed on the first display means by the health care subject is converted into voice data or image data. And reading the first and second data from the server via the communication network, a first terminal that transfers the converted data to the server as the second data via the communication network, A health data collecting apparatus comprising: a second terminal that converts at least the audio data or image data as the second data into audio or image It has been described.

  In the health data collection device of the above-mentioned Patent Document 1, a question at the time of an inquiry is displayed on the first display means provided in the first terminal, and the examinee reads the question and answers by voice or a touch panel. It is configured as follows. In this case, there is a problem that it is often difficult for a low vision person such as an elderly person to read a question.

  On the other hand, in Patent Documents 2 and 3, a medical information processing apparatus placed on the medical staff side and a health management support terminal placed on the user side are connected to a server by a network, and the health management support terminal is By providing voice output means for performing voice output based on voice synthesis or voice data, a question stored in advance and / or a question added by the server is given to the user and returned from the user The information is acquired by an information acquisition means for acquiring information, and the response data is automatically transmitted to the server. The server holds a question list for each user, and adds additional question data to the server. At least user health-related information that is transmitted to the health management support terminal and managed by reply data for each user is accumulated, and in response to a request from the medical information processor A health management support configured to transmit at least the user health-related information to the apparatus and to receive an additional question for each user input from the medical information processing apparatus The system is described.

  In this health management support system, questions during an inquiry are made by voice, and the answers of the examinees are also made by voice. Therefore, since the interview is performed in an interactive manner, it is possible to perform an appropriate interview for a weakly sighted person such as an elderly person.

  FIG. 19 is a flowchart showing the operation of the health management support system described in Patent Document 2. This system is based on conducting 3-4 questions a day and making a round in a week. For example, when the number of interviews is 23, the number of interviews on each day of the week is 4, 4, 3, 3, 3, 3, 3. Steps S101 to S105 in FIG. 19 are preprocessing such as selection of a question when performing one inquiry. The inquiry operation at the time of one inquiry is steps S106 to S116 in FIG.

  First, in steps S106 and 107, a question is read out, and an inquiry word is uttered at a health management support terminal (pet type health management support terminal) on the user side. The waiting time for answering “Yes” or “No” by the user for this inquiry is 10 seconds (S109). If there is no answer after 10 seconds, speak again. This is repeated up to three times (S111). If there is no voice input even after three executions, the answer to that question is “No” (S113).

When the answer is “Yes” (S110), the health care support terminal will announce the question “It may be a symptom of diabetes. If the user's answer to this is “Yes”, it is processed as “Registration is possible” (“Yes” in Step S114), and information of “Server registration is possible” is stored in Step S115. If the answer of the user is “No” in step S114, the information “server registration impossible” is stored in step S115.
In steps S108 to S110, when the user cannot hear the contents of the inquiry, if the user responds “again”, the process returns to step S107 and the contents of the inquiry are again uttered.

  If there is no answer even if the utterance is repeated, please say “Yes” to the pet if the answer is “Yes”, and “No” to the pet if the answer is “No”. If the answer is still not obtained, utter “Do you want to stop the interview?” And if the answer is “Yes” (“Cancel” in step S110), the interview task is terminated. If the answer is “No”, the same question item is spoken again.

Then, it is determined whether or not the inquiry schedule for the day has been completed (S116). If completed, the inquiry task is terminated, and if not completed, the process returns to step S106.
JP 2008-18136 A JP 2003-122846 A JP 2003-122847 A

  However, in the interrogation system of Patent Document 2, it is assumed that the user gives a prescribed response such as “Yes”, “No”, or “Once again” in the dialog processing of Steps S107 to S110. Processing when the user makes an unexpected response is not considered. Therefore, if the user has an unexpected response, it is determined that there is no response in step S109. Therefore, after the interview utterance of the same interview content is repeated three times, the answer is regarded as “no” (S113). That question will end.

  If the user does not understand the operation of such a system, the user repeats an unexpected response, and eventually there is a problem that an accurate inquiry result cannot be obtained. In particular, such a situation is expected to occur in the case of a user who is not good at using a complicated system such as an elderly person.

  Accordingly, an object of the present invention is to provide an automatic inquiry system that can guide a user to make an assumed answer when an answer is made other than an expected answer to an inquiry. It is in.

  A first configuration of an automatic inquiry system according to the present invention is an automatic inquiry system that automatically performs an inquiry to a medical examinee, and a microphone that inputs voice to the medical examinee, and a speaker that outputs audio to the medical examinee, Voice recognition means for converting the voice inputted by the examiner from the microphone into character data and outputting; voice output means for converting the character data to voice and outputting from the speaker; an automatic inquiry system in the interview; Question information storage means for storing question information, which is a question sentence for an examinee from an automatic inquiry system in an inquiry dialogue scenario, which is a progress of the dialogue, and which is a question sentence requesting an answer by a plurality of options, and the question An inquiry result storage means for storing answer information, which is information of a selected answer selected by the examinee for the sentence, in association with the question sentence; and the question information storage The question information stored in the means is sequentially read out, the question reading means for outputting the question sentence from the speaker by the voice output means, and immediately after the question reading means outputs the question sentence, the examinee Answer that discriminates whether answer information input from the microphone and converted into character information by the voice recognition means is an assumed answer for selecting an option assumed for the immediately preceding question sentence, or other unexpected answer Determining means; guide information storage means for storing guide information which is guide sentence information for guiding to make a correct selection answer to the question sentence according to the type of each question sentence choice; and the answer When the determination means determines that the answer information is the assumed answer, the answer stores the answer information in association with the question sentence in the inquiry result storage means When the answer information is determined to be the unexpected answer by the existing means and the answer determination means, the guide information stored in the guide information storage means is read according to the type of option of the question sentence, And a guide output means for outputting the guide information from the speaker by the voice output means.

  With this configuration, when the examinee responds to the question text unexpectedly, the guide output means reads out the guide text for guiding the correct selection answer to the question text, and prompts the response again. . Therefore, even a patient who is not good at using a complicated system can be guided to select a prescribed answer to the interview through dialogue, and the accuracy of the interview result can be improved. .

  Here, the “question text for the examinee” means each question on the questionnaire. The “question text that asks for an answer with a plurality of options” refers to a question text that asks for an answer that selects a plurality of options such as “yes”, “no”, “desired”, and “not desired”. The number of options is not limited to two, and may be three or more options such as “applicable”, “not applicable”, and “neither”.

  In addition, in a second configuration of the automatic inquiry system according to the present invention, in the first configuration, the answer determination unit is configured so that the examinee is connected to the microphone immediately after the voice output unit utters and outputs the question information. Whether the answer information input and converted into character information by the voice recognition means is an assumed answer for selecting an option specified for the immediately preceding question information, an interrupt control instruction, or any other unexpected answer And an interrupt control means for executing an interrupt operation specified by the interrupt control instruction when the answer determination means determines that the answer information is the interrupt control instruction. It is characterized by.

  Here, the “interrupt control instruction” means an instruction for prompting the utterance of a question again such as “again”, an instruction for interrupting the interview such as “stop”, and an inquiry from the beginning such as “from the beginning” This is an instruction to change or interrupt the flow of the question / response loop in the interview, such as an instruction to redo. When an interrupt control instruction is input by the examinee, the interrupt control means executes an interrupt operation specified by the interrupt control instruction. Thereby, it is possible to repeat or interrupt the inquiry according to the will of the examinee.

A third configuration of the automatic inquiry system according to the present invention is an automatic inquiry system that automatically performs an inquiry for a patient,
A microphone for the examinee to input voice,
A speaker that outputs sound to the examinee;
A voice recognition means for converting the voice inputted by the examinee from the microphone into character data and outputting it;
Voice output means for converting character data into voice and outputting from the speaker;
Each information processing unit on the system side in the interview dialogue scenario, which is the progress of the dialogue between the automatic interview system and the examinee in the interview, is called a processing item, and each processing item that accompanies the progress of the dialog process according to the interview dialog scenario When the transition relationship between them is called a net,
A record is assigned to each of the processing items, and each record specifies input net information for specifying an input net on the transition side to the processing item and an output net on the side of transition from the processing item. A processing item table storage means for storing a processing item table storing a corpus that describes output net information and utterance contents or receiving keywords set as necessary for the processing item;
A record is assigned to each of the nets, and each record has a net state table storage for storing a net state table in which the net ID of the net and a net state flag indicating the active state of the net are stored. Means,
An interrogation result storage means for memorizing interrogation results;
With
The processing item table is at least:
As the corpus, a start keyword for starting the interview dialogue scenario is stored, and as the output net information, a net of a net that transitions to a processing item that performs an utterance process of the first question sentence or guide sentence in the interview dialogue scenario A start information record group in which an ID is stored;
As the corpus, a question sentence that is a question sentence for the examinee and that asks for answers by a plurality of options is stored, and the input net information is a network that is connected to a processing item that is a transition source to the question sentence in the inquiry dialogue scenario. A group of question information records in which the net ID of the net connected to the processing item for determining the answer of the examinee to the question sentence in the inquiry dialogue scenario is stored as the output net information;
As the input net information, a net ID of the input net connected to any of the question sentence processing items is stored, and as the corpus, an answer that is a keyword of a selected answer selected by the examinee for the question sentence A plurality of keywords are stored and, as the output net information, a net ID of the output net selected according to the answer keyword and a net ID of the output net selected for an unexpected answer are stored Information records,
As the corpus, a guide sentence and other guide sentences for guiding a correct selection answer to the question sentence corresponding to each question sentence and other guide sentences are stored, and the guide information in the inquiry dialogue scenario is used as the input net information. The net ID of the net connected to the process item that is the transition source to the sentence is stored, and the net ID of the net that is connected to the process item that is the transition destination from the guide sentence in the inquiry dialogue scenario is stored as the output net information. Guide information record group,
Including
When the character data is input from the voice recognition unit, the corpus that matches or is included in the character data among the records of the start information record group is searched, and the output net of the record extracted by the search Scenario start means for obtaining a net ID and activating a net state flag corresponding to the net ID obtained in the net state table storage means;
If all the records of the question information record group are sequentially selected and the net status flag of the net status table corresponding to the input net of the selected record is active, the question text of the selected record is A question that is output by the output means, deactivates the net status flag corresponding to the input net of the selected record of the net status table, and activates the net status flag of the output net of the selected record Sentence output means;
When the character data is input from the voice recognition means, the record of the net status flag of the net status table for the input net is searched for among the records of the answer information record group, and the search When a result record is extracted, it is determined whether any one of the answer keywords of the record matches or is included in the character data. If there is the answer keyword that matches or is included in the character data, The net ID for the answer keyword is acquired, the net status flag corresponding to the net ID acquired in the net status table storage means is activated, the net status flag for the input net is deactivated, and the answer While storing keyword selection results in the inquiry result storage means, If there is no answer keyword that matches or is included in the written data, the net ID for the unexpected answer is acquired, and the net status flag corresponding to the net ID acquired in the net status table storage means is activated. And answer determination means for deactivating the net status flag for the input net;
If all the records of the guide information record group are sequentially selected and the net status flag of the net status table corresponding to the input net of the selected record is active, the guide sentence of the selected record is converted to the voice. A guide for outputting by the output means, deactivating the net status flag corresponding to the input net of the selected record of the net status table, and activating the net status flag of the output net of the selected record And a sentence output means.

In addition, in the fourth configuration of the automatic inquiry system according to the present invention, in the third configuration, the answer information record group is connected to any one of the question sentence processing items as the input net information. A net ID of an input net is stored, a plurality of the answer keywords and an interrupt control keyword for instructing interrupt control are stored as the corpus, and the answer keyword and the interrupt are output as the output net information. The net ID of the output net selected according to the control keyword and the net ID of the output net selected for the unexpected answer are stored,
The answer discriminating means is a record in which the net status flag of the net status table for the input net is active in each record of the answer information record group when the character data is inputted from the voice recognition means. When a search result record is extracted, it is determined whether or not any of the answer keywords in the record matches or is included in the character data,
If there is an answer keyword that matches or is included in the character data, the net ID corresponding to the answer keyword or the interrupt control keyword is obtained, and the net corresponding to the net ID obtained in the net state table storage means Activate the status flag, deactivate the net status flag for the input net, and store the selection result of the answer keyword in the inquiry result storage means,
If there is the interrupt control keyword that matches or is included in the character data, the net ID for the interrupt control keyword is acquired, and the net status flag corresponding to the net ID acquired in the net status table storage means And deactivating the net status flag for the input net,
If there is no answer keyword that matches or is included in the character data, the net ID for the unexpected answer is acquired, and the net state flag corresponding to the net ID acquired in the net state table storage means is activated. In addition, the net status flag for the input net is deactivated.

  As described above, according to the present invention, when the examinee makes an unexpected response to the question text, the guide output means guides the user to make a correct selection answer to the question text. By reading out the sentence and prompting a response again, it is possible to guide the user to select a prescribed answer to the interview through dialogue. Therefore, it is possible to obtain a highly accurate inquiry result even for a patient who is not good at using a complicated system such as an elderly person.

  The best mode for carrying out the present invention will be described below with reference to the drawings.

  FIG. 1 is a diagram illustrating a configuration of an automatic inquiry system according to Embodiment 1 of the present invention. The automatic inquiry system 1 has a configuration in which a management server 2, a diagnostician side computer 3, and a examinee side terminal device 4 are connected to be communicable via a network 5. The diagnostician computer 3 is a computer used by each doctor who conducts an inquiry to view and output the inquiry results. The examinee side terminal device 4 is a terminal used by the examinee who receives the inquiry at the time of the inquiry. The examinee side terminal device 4 can also use a dedicated device as shown in FIG. 1 in addition to a normal personal computer. In FIG. 1, one each of the diagnoser-side computer 3 and the examinee-side terminal device 4 are illustrated, but there may be a plurality of these.

  The examinee side terminal device 4 includes a speaker 6, a microphone 7, and a touch panel 8. The speaker 6 and the microphone 7 are used for dialogue with the examinee. The touch panel 8 is provided for displaying a message for the examinee and receiving an input from the examinee.

  FIG. 2 is a diagram illustrating a functional configuration of the automatic inquiry system 1 of FIG. In FIG. 2, the management server 2, the examiner side computer 3, the examinee side terminal device 4, the network 5, the speaker 6, the microphone 7, and the touch panel 8 correspond to the same reference numerals in FIG.

  In FIG. 2, the management server 2 includes a question information storage unit 11, an inquiry result storage unit 12, and a communication interface (communication I / F) 13. The question information storage unit 11 and the inquiry result storage unit 12 are realized by a storage device built in the management server 2.

  The question information storage unit 11 stores question information that is a question text for a question to the examinee and information on a question text that asks for answers by a plurality of options. The inquiry result storage unit 12 stores answer information, which is information of a selected answer selected by the examinee with respect to the question sentence of the questionnaire, in association with the question sentence. The communication I / F 13 is a communication interface for the management server 2 to communicate with the diagnostician side computer 3 and the examinee side terminal device 4 via the network.

  The diagnostician side computer 3 is a computer for the diagnostician to refer to the inquiry result.

  The examinee side terminal device 4 includes a speaker 6, a microphone 7, a touch panel 8, a voice output unit 14, a voice recognition unit 15, a question group selection unit 16, a question reading unit 17, an answer determination unit 18, an answer storage unit 19, and an interrupt. A control unit 20, a guide information storage unit 21, a guide output unit 22, and a communication interface 23 are provided.

  The voice output unit 14 converts the character data into voice and outputs it from the speaker 6. The voice recognition unit 15 converts voice input from the microphone 7 by the examinee into character data and outputs the text data. The question group selection unit 16 selects an inquiry dialogue scenario, which is the progress of dialogue between the automatic inquiry system and the examinee, in accordance with the instruction of the examinee. The question reading unit 17 sequentially reads the question information stored in the question information storage unit 11 via the network 5, and the voice output unit 14 utters and outputs the question sentence from the speaker 6. Immediately after the question text reading unit 17 utters and outputs the question text, the answer discriminating unit 18 inputs the answer information input from the microphone 7 and converted into character information by the voice recognition unit 15 to the previous question text. It is discriminated whether it is an assumed answer for selecting an assumed option, an interrupt control instruction, or any other unexpected answer. Here, the “assumed answer” means a selected answer assumed for the question sentence. When the answer determining unit 18 determines that the answer information is an assumed answer, the answer storing unit 19 stores the answer information in the inquiry result storage unit 12 in association with the question text. When the answer determination unit 18 determines that the answer information is an interrupt control instruction, the interrupt control unit 20 executes an interrupt operation specified by the interrupt control instruction.

  The guide information storage unit 21 stores guide information that is guide sentence information for guiding the user to make a correct selection answer to the question sentence according to the type of option of each question sentence. When the answer determination unit 18 determines that the answer information is an unexpected answer, the guide output unit 22 reads the guide information stored in the guide information storage unit 21 according to the type of question sentence option, The guide information is output from the speaker 6 by the voice output unit 14.

  The communication interface 23 is a communication interface for the examinee side terminal device 4 to communicate with the diagnoser side computer 3 and the management server 2 via the network.

  FIG. 3 is a diagram illustrating a data structure of a question information record group stored in the question information storage unit 11. The question information record group has six fields: a “question ID” field, a “question group” field, a “question sentence” field, an “assumed answer 1” field, an “assumed answer 2” field, and an “assumed answer 3” field. . The example in FIG. 3 shows a case where there are a maximum of three assumed answers, but when there are four or more assumed answers, the “assumed answer 4” field, the “assumed answer 5” field,... Just increase it.

  The “question ID” field is a field in which an ID unique to the question sentence is stored. The “question group” field is a field in which information for identifying a question group to which the question sentence belongs is stored. “Question group” means a set of questions included in one questionnaire. For example, a “stress check questionnaire” is a “stress check” question group, and a “health questionnaire” is a “health questionnaire”. “One-to-one correspondence with the questionnaire, such as a group of questions”. The “question text” field stores a question text for voice output or screen output. In the “assumed answer n” field (n = 1, 2,...), Assumed answers to the question sentence are stored one by one.

  FIG. 4 is a diagram illustrating the relationship between the data structure of the inquiry result table and the examinee table stored in the inquiry result storage unit 12 and other tables. FIG. 4A shows the structure of the inquiry result table, and FIG. 4B shows the structure of the examinee table. FIG. 4C shows a relationship between the inquiry result table, the examinee table, and the question information record group.

  The examinee table has three fields of “examinee ID”, “name”, and “name kana”. In the “checker ID” field, a checker ID assigned to each checker is stored. In the “name” field, the name of each examinee is stored. In the “Name Kana” field, the reading of the name of each examinee is stored.

  The inquiry result table has four fields of “inquiry result ID”, “examinee ID”, “question ID”, and “answer number”. In the “interview result ID” field, an ID (interview result ID) uniquely assigned to each interview result is stored. The “checker ID” field stores the checker ID of the checker who has made the inquiry. The “question ID” field stores the question ID of the question sentence for which the inquiry was made. In the “answer number” field, the number (answer number) of the selected answer answered by the examinee with respect to the question sentence is stored. In this case, the answer number is “1” when “assumed answer 1” of the question information record group is selected, the answer number is “2” when “assumed answer 2” is selected, and so on. Are allocated as follows.

  The relationship between the inquiry result table, the examinee table, and the question information record group is as shown in FIG. There is a one-to-many relationship between the “examinee ID” in the examinee table and the “examinee ID” in the inquiry result table. The “question ID” in the question information record group and the “question ID” in the inquiry result table have a one-to-many relationship.

  About the automatic inquiry system of a present Example comprised as mentioned above, the operation | movement is demonstrated below.

  FIG. 5 is a flowchart showing the operation of the automatic inquiry system 1 according to the first embodiment of the present invention.

  First, the examinee side terminal device 4 displays the name and examinee ID input screen as shown in FIG. 6 on the screen of the touch panel 8. The examinee inputs the name and the examinee ID from the touch panel 8 in the examinee side terminal device 4 (S1). The examinee side terminal device 4 transmits the input name and examinee ID to the management server 2. The patient ID is not essential.

  The management server 2 identifies the examinee based on the name and the examinee ID, and types of interview dialogue scenarios of the interview conducted for the examinee (for example, “stress check”, “health questionnaire”, “HAD” The scale ”,“ TAS-20 ”, etc.) are transmitted to the terminal device 4 on the examinee side. The question group selection unit 16 of the examinee-side terminal device 4 displays an inquiry form selection screen as shown in FIG. 7 on the screen of the touch panel 8. The examinee selects an inquiry form (question group) to receive an inquiry from this selection screen (S2). When the questionnaire is selected, the examinee side terminal device 4 proceeds to the operation of the next step S3. Here, it waits until an inquiry form is selected, or when “return to name input” is selected, the terminal unit 4 on the examinee side closes the selection screen of the inquiry form and returns to step S1.

  When the specific information of the selected questionnaire (question group) is received, the guide output unit 22 of the terminal device 4 on the examinee first outputs a guide sentence explaining how to proceed with the interview to the voice output unit 14, The output unit 14 utters and outputs the guide sentence to the examinee through the speaker 6 (S3). Thereby, the examinee can understand the operation method of the automatic inquiry system 1.

  As this “guide sentence”, for example, “(Stress check) will start. Since we will ask (13) questions, please answer any one (applies, does not apply, cannot say either). I will repeat the answer here, so if the answer is different, say "Back" to the microphone or press the "Back" button on the touch panel. Is output. In the above guide sentence, the part surrounded by () is a part that is appropriately changed according to the question group.

  Next, the question reading unit 17 sequentially reads the question information stored in the question information storage unit 11 from the question information record group via the network 5. In this case, since the question group is specified, the question information records belonging to the specified question group are read in ascending order of the question ID. Then, the question reading unit 17 utters and outputs the question text of the read question information record from the speaker 6 by the voice output unit 14 and also displays the question information on the touch panel 8 based on the assumed answer of the question information record in FIGS. A question guide as shown is displayed (S4).

  Here, the question sentence is a question sentence that asks the examiner for a selection answer from a plurality of options, and the examiner selects and answers one option from the presented options. There are two or more options. For example, when there are two options, the touch panel 8 displays a screen as shown in FIG. 8, and when there are three options, the touch panel 8 displays a screen as shown in FIG. .

  8 and 9, a question sentence (for example, a part that is displayed as “Nervous or is it said by a person” in FIG. 8) is displayed in the upper frame of the screen display, and is displayed in the center. Is displayed as an answer sentence that can be selected as an option (for example, a portion where “Yes” and “No” are displayed in FIG. 8) and a selection button (for example, a circular button surrounding “Yes” and “No” in FIG. 8). The When the examinee touches these selection buttons with a finger, an answer to the question sentence is input to the touch panel 8. On the other hand, since the same question sentence (for example, the question sentence “Is it nervous or is it said by a person?” In FIG. 8) is spoken from the speaker 6, the contents of the question are also given to the examinees with low vision. Can communicate. The examinee can also input a choice by voice from the microphone 7 in response to the question sentence.

  Further, at the bottom of the screen display, an interrupt control instruction sentence (for example, “return to previous”, “question”, “from the beginning” sentence in FIGS. 8 and 9) and an interrupt control instruction button (for example, FIG. 8 and 9, a straight band button surrounding a sentence “return to the front”, “question a question”, and “from the beginning” is displayed. For example, in FIGS. 8 and 9, when the examinee touches the interrupt control instruction button with a finger, the interrupt control instruction is input to the touch panel 8.

  Next, the answer discriminating unit 18 waits for an answer to the question text from the examinee, and discriminates the content when the answer is inputted (S5). When an answer is input from the microphone 7, the voice recognition unit 15 converts the input voice into a text format and inputs it to the answer determination unit 18. The answer determination unit 18 uses a specific keyword in the answer sentence. Is discriminated by determining whether or not is included.

  When the answer is an assumed answer assumed for the question sentence, the answer determination unit 18 transitions to any of steps S6, S7, and S8 according to the contents of the answer.

  When the content of the assumed answer is affirmative (“Yes”), the answer storage unit 19 creates a new record in the inquiry result table of the inquiry result storage unit 19 via the network 5, and adds a new inquiry result ID to the record. And the answer number (in this case, “1”) indicating affirmation (“Yes”) is stored together with the examinee ID and the question ID (S6). When the content of the assumed answer is negative (“No”), the answer storage unit 19 creates a new record in the interview result table of the interview result storage unit 19 via the network 5 and adds a new interview result ID to the record. And the answer number (in this case, “2”) indicating negative (“No”) is stored together with the examinee ID and the question ID (S7). When the content of the assumed answer is other than affirmative or negative (“cannot be said”), the answer storage unit 19 creates a new record in the interview result table of the interview result storage unit 19 via the network 5, and the record A new inquiry result ID is stored, and an answer number (in this case, “3”) representing indefinite (“neither”) is stored together with the examinee ID and question ID (S8).

  After this step S6, S7, S8, if this question sentence is not the final question (S9), the question sentence reading unit 17 selects the next question and returns to the process of step S4. On the other hand, in the case of the final question (S9), the guide output unit 22 outputs a guide sentence to the effect that the question has been completed from the speaker 6 via the voice output unit 14, and on the screen of the touch panel 8 as shown in FIG. An end screen is output (S11), and the process ends.

  In step S5, when the answer is an interrupt control instruction, processing according to the interrupt control instruction is performed.

  When an interrupt control instruction of “quick a question” is input, the process returns to step S4 in order to output the same question sentence again, and the question reading unit 17 sends the same question sentence from the speaker 6 via the voice output unit 14. Is uttered.

  When the “return to previous” interrupt control instruction is input, the question reading unit 17 reads the previously read question information record from the question information record group to output the previous question sentence (S13), and proceeds to step S4. Returning, the question reading unit 17 utters and outputs the previous question sentence from the speaker 6 via the voice output unit 14, and outputs the previous question to the touch panel 8.

  When an interrupt control instruction “from the beginning” is input, the question reading unit 17 changes the first question information record (with the smallest question ID) of the current question group to be output from the first question sentence. Reading from the group (S14), returning to step S4, the question reading unit 17 utters and outputs the first question sentence from the speaker 6 via the voice output unit 14, and outputs the first question to the touch panel 8.

  In step S5, when the answer is an unexpected answer other than the assumed answer and the interrupt control instruction, the guide output unit 22 is stored in the guide information storage unit 21 according to the type of option of the question sentence. The guide information is read out, and the guide information is uttered from the speaker 6 by the voice output unit 14 (S15). Then, in order to output the same question sentence again, the process returns to step S <b> 4, and the question reading unit 17 outputs the same question sentence from the speaker 6 via the voice output unit 14. As described above, when the examinee makes an unexpected response to the question text, the guide output unit 22 reads out the guide text for guiding the correct selection answer to the question text, and sends a response again. In order to promote, even a patient who is not good at using the machine can be guided to select the expected answer to the interview correctly through dialogue, and the accuracy of the interview result can be improved. .

  In the present embodiment, the question information storage unit 11 and the inquiry result storage unit 12 are provided in the management server 2 and connected to the examinee side terminal device 4 via the network 5, but the question information storage unit 11 and the inquiry result storage unit 12 may be provided in the examinee side terminal device 4.

  In the present embodiment, an example is shown in which the inquiry result is stored in the inquiry result table stored in the inquiry result storage unit 12 in the form of a relational database. However, in the present invention, the method for saving the inquiry result is not limited to this method. The interview number and the answer ("Yes", "No", etc.) for each interview content are written and stored in the interview result storage unit 12 in the interview form file created for each interview for the examinee. You may do it.

  FIG. 11 is a diagram illustrating a functional configuration of the automatic inquiry system according to the second embodiment. In this embodiment, the same system as in FIG. 1 is used as the hardware configuration.

  The automatic inquiry system 1 according to this embodiment includes a management server 2 and a receiver side terminal 4 connected via a network. The management server 2 includes an inquiry result storage unit 12 in which an inquiry result table that stores inquiry results, a recipient table that stores recipient information, and a question information table that stores question information are stored. . It is assumed that the inquiry result table, the recipient table, and the question information table have the same data structure as FIG.

  The receiver side terminal 4 includes a speaker 6, a microphone 7, a touch panel 8, and a computer 30. The speaker 6, the microphone 7, and the touch panel 8 are the same as those in the first embodiment.

  The computer 30 executes a program for an automatic inquiry system, thereby realizing each functional block as shown in FIG. That is, the computer 30 includes a recognition dictionary storage unit 31, a voice recognition unit 32, a process item table storage unit 33, a net state table storage unit 34, a scenario start unit 35, an answer determination unit 36, a question sentence output unit 37, and an answer output unit. 38, an inquiry result storage module 39, a guide sentence output unit 40, a voice output module 41, a connection point / end point processing unit 44, a screen display module 45, and an input recognition module 46.

  The recognition dictionary storage unit 31 stores a recognition dictionary used for speech recognition. The voice recognition unit 32 converts a voice signal input from the microphone 7 into text data using a recognition dictionary. The voice output module 41 converts character data into voice and outputs it from the speaker 6.

  The process item table storage unit 33 stores a process item table. The net state table storage unit 34 stores a net state table. The processing item table and the net state table will be described in detail later.

  The processing item table is composed of a set of a large number of records as shown in FIGS. 14 and 15, and these records include a start information record group, a question information record group, an answer information record group, a guide. The information record group, the answer output record group, the connection point record group, and the end point record group are classified into respective record groups (a set of records). The meaning of each record group will be described in detail later.

  The scenario start unit 35 and the answer determination unit 36 form part of the recognition information input unit 42. The question sentence output unit 37, the answer output unit 38, the guide sentence output unit 40, and the connection point / end point processing unit 44 constitute a part of the selection execution unit 43. The functions of these units will be described later.

  The inquiry result storage module 39 is a module for storing answers to the interview of the examinee in an inquiry result table stored in the inquiry result storage unit 12.

  The screen display module 45 displays an image on the screen of the touch panel 8. When an input is made by touching the screen of the touch panel 8 with a finger, the input recognition module 46 recognizes the input type from the contact position on the screen and outputs a recognition result.

  Next, an inquiry dialogue scenario, which is the progress of dialogue between the automatic inquiry system 1 and the examinee when the automatic inquiry system 1 in this embodiment performs an inquiry, will be described.

  Dialogue processing according to the interview dialogue scenario as seen from the side of the automatic inquiry system 1 is represented by transition diagrams as shown in FIGS. FIG. 12 shows a transition diagram of the interview dialogue scenario of the stress check questionnaire, health questionnaire questionnaire, HDA scale questionnaire, and TAS-20 questionnaire. The part indicated by the dotted line is the omitted part. FIG. 13 shows an entire transition diagram of the interview dialog scenario of the stress check questionnaire.

  Here, each information processing unit of the automatic inquiry system 1 in the inquiry dialogue scenario is referred to as a “processing item”, and a transition relation between the processing items accompanying the progress of the dialogue process in accordance with the inquiry dialogue scenario is referred to as a “net”. In FIG. 12 and FIG. 13, each processing item is indicated by a square, a rhombus, a rounded square, a point, or a fringe shape (a pentagon in which an isosceles triangle is joined to one side of the rectangle from the bottom side). A line connecting each processing item represents a net.

  “FCT” represents a processing item for extracting a specified keyword described in an XML (Extensible Markup Language) format for speech recognition from input speech. “<FCT>” represents a corpus (keyword) for entering the FCT.

  “TTS” (Text to Speech) represents a text-to-speech process, that is, a process item for converting text data into speech and outputting it from the speaker 6. “<SP>” represents the content of the text read out in the TTS. “QB” represents a processing item for extracting a speech word from text data output from the speech recognition unit 32 using a specified XML and branching the flow of dialogue according to the extracted speech word. Questions are described by using both TTS and QB. For example, in TTS2, in “QB1” that utters “Question 1. Frustrated even a little” and recognizes the answer voice of the receiver to this, the expected answer method (in this case, “applicable”, “not applicable”) ”,“ Neither ”,“ Back to the previous ”,“ From the beginning ”,“ End ”) are specified, thereby realizing the process of recognizing the answer to the question.

  “OUT” represents a processing item for outputting and storing the answer of the examinee in the inquiry result table. “JA” represents a connection point of the net. “TRM” represents a processing item of termination processing. That is, this inquiry dialogue scenario is completed in TRM.

  Also, the number attached to each line (net) represents the net ID of the net. Hereinafter, a net whose net ID is i (i is an integer) is referred to as NETi.

  14 and 15 are diagrams illustrating an example of the processing item table stored in the processing item table storage unit 33. The processing item tables shown in FIG. 14 and FIG. 15 are for the interview dialog scenario of the stress check questionnaire shown in FIG. The process item table is configured by a set of a large number of records, and each record is configured by fields of “process item”, “input net”, “output net”, and “corpus”.

  The “process item” field stores a specific symbol (or ID) of the process item corresponding to the record, such as TTSi or FCTi. The “input net” field stores the net ID of the net (input net) on the transition side to the processing item. The “output net” field stores the net ID of the net (output net) on the transition side from the processing item. The number of input nets and output nets may be one, plural, or none. In the “corpus” field, a corpus that describes utterance contents or reception keywords set as necessary for the processing item is stored.

  Each record of the processing item table is classified into a start information record group, a question information record group, an answer information record group, a guide information record group, an answer output record group, a connection point record group, and an end point record group. .

  In the start information record group, a start keyword for starting the interview dialogue scenario is stored in the “corpus” field, and the first question sentence or guide sentence utterance processing in the interview dialog scenario is stored in the “output net” field. This is a set of records in which the net IDs of output nets transitioning to items are stored. Each record in the start information record group corresponds to an FCT process item.

  The question information record group is a set of records corresponding to processing items in which the automatic inquiry system 1 utters a question sentence to the examinee. A part of the TTS process items corresponds to each record of the question information record group. In each record of the question information record group, a question sentence that is a question sentence for the examinee and asks for an answer by a plurality of options is stored in the “corpus” field, and the question sentence in the question dialogue dialog scenario is stored in the “input net” field. The net ID of the input net that is connected to the process item that is the source of the transition to is stored, and the net of the output net that is connected to the process item that determines the response of the examinee to the question text in the question dialog scenario in the “output net” field ID is stored.

  The answer information record group is a set of records corresponding to the process items that branch the process flow in accordance with the examinee's answer to the question. Each record in the answer information record group corresponds to a QB process item. Each record of the answer information record group stores the net ID of the input net connected to the processing item of any question text in the “input net” field, and the examinee for the question text in the “corpus” field. A plurality of response keywords and interrupt control keywords for instructing interrupt control are stored, and the “output net” field is selected according to the response keyword and interrupt control keyword. The net ID of the output net and the net ID of the output net selected for the unexpected answer are stored. Here, the “unexpected answer” refers to an answer other than the answer including the answer keyword assumed for the question (assumed answer) and the answer including the interrupt control keyword.

  The guide information record group is a set of records corresponding to processing items in which the automatic inquiry system 1 utters a guide sentence to the examinee. A part of the TTS process item corresponds to each record of the guide information record group. Each record of the guide information record group stores, in the “corpus” field, a guide sentence and other guide sentences that guide the user to make a correct selection answer corresponding to each question sentence. ”Field stores the net ID of the input net connected to the processing item of the transition to the guide sentence in the interview dialogue scenario, and the“ output net ”field stores the processing of the transition destination from the guide sentence in the interview dialogue scenario. Stores the net ID of the output net connected to the item.

  The answer output record group is a record group corresponding to a processing item that outputs and stores the contents of the examinee's answer to the question in the inquiry result table. An OUT process item corresponds to each record in the answer output record group. Each record of the answer output record group stores the net ID of the input net that transitions from any QB process item in the “input net” field, and the output that transitions to the next TTS process item in the “output net” field. The net ID of the net is stored.

  The connection point record group is a set of records corresponding to processing items representing the connection point of the net. A JA processing item corresponds to each record in the connection point record group. At each connection point, no input / output process is performed, and only a process for changing the net state is performed. That is, at the connection point, when the input net is in an active state, processing is performed in which the output net is in an active state and the input state is in an inactive state.

  The end record group is a set of records corresponding to the process items of the end process. Each record of the end point record group corresponds to a TRM process item.

  FIG. 16 is a diagram illustrating an example of a net state table stored in the net state table storage unit 34. The net state table shown in FIG. 16 is for the interview dialogue scenario of the stress check questionnaire shown in FIG. The net state table is composed of a set of a large number of records, and each record is composed of fields of “net”, “net state”, and “change flag”.

  The “net” field stores the net ID of the net corresponding to the record. The “net status” field stores a net status flag indicating the active status of the net. The net state flag takes a value of “0” or “1”, where “0” represents an inactive state and “1” represents an active state. The “change flag” field stores a change flag indicating whether or not the net state has been changed. The change flag takes a value of “0” or “1”, where “0” indicates no change and “1” indicates a change.

  The inquiry dialog processing operation of the automatic inquiry system 1 according to the present embodiment configured as described above will be described below. In this embodiment, the part where the examiner inputs the name and the examinee ID of the examinee is the same as that of the first embodiment, and the description of this part is omitted. Accordingly, in the following description, it is assumed that the examinee ID of the examinee is already set. Moreover, in the automatic inquiry system 1 of the present embodiment, input / output using a touch panel is possible in addition to voice input / output. However, for the sake of simplicity, the following description will be limited to voice input / output processing. .

(1) Operations Related to Recognition Information Input Processing FIG. 17 is a flowchart showing operations of the speech recognition unit 32 and the recognition information input unit 42. The voice recognition unit 32 and the recognition information input unit 42 operate independently of each other.

(1-1) Operation of Voice Recognition Unit 32 First, in step S21, the voice recognition unit 32 detects whether or not voice is input from the microphone 7. If no voice is input, step S21 is repeated again.

  When speech is input, in step S22, the speech recognition unit 32 refers to the recognition dictionary stored in the recognition dictionary storage unit 31, converts the input speech signal into text data, and recognizes the information input unit 42. Output to. Then, the process returns to step S21.

(1-2) Operation of Recognition Information Input Unit 42 First, in step S31, the recognition information input unit 42 selects the processing item table stored in the processing item table storage unit 33 in order from the first processing item record. . When selection has been completed up to the last process item record, selection is again made in order from the first process item record.

  Next, in step S <b> 32, the scenario start unit 35 checks whether or not the selected process item record is an FCT process item (belongs to the start information record group). If it is an FCT process item, the process proceeds to step S33, and if not, the process proceeds to step S36.

  In step S <b> 33, the scenario start unit 35 checks whether or not speech text data has been input from the speech recognition unit 32. If there is no input of speech text data, the process returns to step S31, and if there is input of speech text data, the process proceeds to the next step S34.

  In step S34, the scenario start unit 35 checks whether the speech text data matches the corpus of the selected FCT process item record or whether the speech text data includes the corpus. If they match or are included, the process proceeds to the next step S35, and if not, the process returns to step S31.

  In step S35, the scenario start unit 35 refers to the selected FCT process item record “output net” field, and acquires the net ID of the output net (hereinafter referred to as “output net ID”). Then, in the net state table stored in the net state table storage unit 34, the net state record whose net ID matches the output net ID is selected, the net state of the selected net state record is activated, and the change flag is set to “ There is a change (1) ". Also, an inquiry form ID corresponding to the selected inquiry dialogue scenario is acquired. Then, the process returns to step S31.

  On the other hand, if the process item record selected in step S32 is not an FCT process item, in step S36, the answer determination unit 36 determines whether the selected process item record belongs to the QB process item (belongs to the answer information record group). Check for no. If it is a QB process item, the process proceeds to the next step S37, and if not, the process returns to step S31.

  In step S37, the answer determination unit 36 acquires the net ID of the input net (hereinafter referred to as “input net ID”) from the “input net” field of the selected QB process item record. Then, in the net state table stored in the net state table storage unit 34, a net state record whose net ID matches the input net ID is selected, and its active state flag is checked. If the active state flag is the active state (1), the process proceeds to the next step S38. If the active state flag is the inactive state (0), the process returns to step S31.

  In step S <b> 38, the answer determination unit 36 checks whether or not speech text data has been input from the speech recognition unit 32. If there is no input of speech text data, the process returns to step S31, and if there is input of speech text data, the process proceeds to the next step S39.

  In step S39, the answer determination unit 36 checks whether the voice text data matches one of the corpuses of the selected QB process item record or whether the voice text data includes the corpus. If they match or are included, the process proceeds to the next step S40, and if not, the process returns to step S31.

  In step S40, the answer determination unit 36 selects the net ID of the output net corresponding to the corpus that matches (or is included in) the voice text data from the net IDs in the “output net” field.

  In step S41, the answer determination unit 36 selects a net state record whose net ID matches the output net ID in the net state table stored in the net state table storage unit 34, and the net state of the selected net state record. Is activated (1), and the change flag is set to “changed (1)”. Further, in the net state table stored in the net state table storage unit 34, a net state record whose input net ID and net ID match is selected, the net state of the selected net state record is deactivated (0), The change flag is set to “no change (0)”. Then, the process returns to step S31.

(2) Operation Related to Selection Execution Processing FIG. 18 is a flowchart showing the operation of the selection execution unit 43. The selection executing unit 43 operates in parallel with the voice recognition unit 32 and the recognition information input unit 42.

  First, in step S51, the selection execution unit 43 selects the process item table stored in the process item table storage unit 33 in order from the first process item record. When selection has been completed up to the last process item record, selection is again made in order from the first process item record.

  In step S52, the selection execution unit 43 acquires the input net ID stored in the “input net” field of the selected process item record. In the net state table stored in the net state table storage unit 34, a net state record whose net ID matches the input net ID is selected, and whether or not the net state of the selected net state record is the active state (1). Check. In the active state (1), the process proceeds to the next step S53, and in the inactive state (0), the process returns to step S51.

  In step S53, the selection execution unit 43 checks whether or not the selected process item record is a TTS process item (belongs to a question information record group or a guide information record group). If it is a TTS process item, the process proceeds to step S53, and if not, the process proceeds to step S55.

  In step S <b> 54, the selection execution unit 43 acquires the corpus of the selected TTS process item record, and outputs the utterance sentence described in the corpus from the speaker 6 via the voice output module 41.

  Here, when the selected TTS process item record belongs to the question information record group, the question sentence output unit 37 outputs the question sentence described in the corpus by voice. When the selected TTS process item record belongs to the guide information record group, the guide sentence output unit 40 outputs a guide sentence described in the corpus by voice. Since the question sentence output unit 37 and the guide sentence output unit 40 are functionally the same except for the utterance contents, they can be configured by one module.

  In step S55, the selection execution unit 43 selects a net state record whose net ID matches the output net ID of the selected processing item record in the net state table stored in the net state table storage unit 34, and selects the selected net state record. The net state of the net state record is activated (1), and the change flag is set to “changed (1)”. Further, in the net state table stored in the net state table storage unit 34, a net state record whose net ID matches the input net ID of the selected process item record is selected, and the net state of the selected net state record is displayed. It is deactivated (0), and the change flag is set to “no change (0)”. Then, the process returns to step S31.

  In step S53, if the selected process item record is not a TTS process item, in step S56, the selection execution unit 43 determines whether the selected process item record is an OUT process item (belongs to an answer output record group). Inspect. If it is an OUT process item, the process proceeds to step S57, and if not, the process proceeds to step S58.

  In step S57, the answer output unit 38 stores the answer number of the answer to the question sentence in the inquiry result table via the inquiry result storage module 39. At this time, the inquiry result storage module 39 stores the answer number in association with the examinee ID of the examinee and the question ID of the question sentence. And it changes to the above-mentioned step S55.

  In step S56, if the selected process item record is not an OUT process item, in step S58, the connection point / end point processing unit 44 determines that the selected process item record belongs to the JA process item (the connection point record group). ) Or not. If it is a JA processing item, the process proceeds to step S55, and if not, the process proceeds to step S59.

  If the selected process item record is not a JA process item in step S58, in step S59, the connection point / end point processing unit 44 determines that the selected process item record is an OUT process item (belongs to the end point record group). Inspect whether or not. If it is an OUT process item, the net state of the input net is deactivated and the inquiry dialogue process is terminated. Otherwise, the process returns to step S51.

  As described above, the inquiry dialogue process in the automatic inquiry system 1 is executed.

  Finally, as an example, a dialogue check process of the stress check questionnaire in the automatic inquiry system 1 will be described with reference to FIG.

(Scenario step 1)
First, a voice “stress check” is input from the microphone 7 by the examinee. This voice is converted into text data by the voice recognition unit 32 and input to the recognition information input unit 42. The recognition information input unit 42 cyclically selects each processing item record in the processing item table, and when the text data is input, the processing item record FCT1 having a corpus that matches the keyword “stress check”. Select. Since this process item record FCT1 belongs to the start information record group, the scenario start unit 35 activates the output net NET1 of FCT1.

(Scenario step 2)
On the other hand, the selection execution unit 43 cyclically selects each processing item record in the processing item table in parallel with the recognition information input unit 42. When NET1 is activated, the recognition information input unit 42 selects NET1. Select the process item record JA1 as the input net. Since this process item record JA1 belongs to the connection point record group, the connection point / end point processing unit 44 activates the output net NET2 of JA1 and deactivates the input net NET1.

(Scenario step 3)
Next, when NET2 is activated, the selection execution unit 43 selects a process item record TTS1 having NET2 as an input net. Since this processing item record TTS1 belongs to the guide information record group, the guide sentence output unit 40 starts the corpus “stress check of TTS1 via the voice output module 41. Since 13 questions will be asked from now on, it applies. Please answer with one of them, which doesn't apply.If you answer wrong, say "Back to the front" or "Previous" on the touch panel. Press the “Return” button. Then, the output net NET3 is activated and the input net NET2 is deactivated.

(Scenario step 4)
Next, when NET3 is activated, the selection execution unit 43 selects a process item record JA2 having NET3 as an input net. Since this process item record JA2 belongs to the connection point record group, the connection point / end point processing unit 44 activates the output net NET4 of JA2 and deactivates the input net NET3.

(Scenario step 5)
Next, when NET4 is activated, the selection execution unit 43 selects a process item record TTS2 having NET4 as an input net. Since this processing item record TTS2 belongs to the question information record group, the question sentence output unit 37 outputs the corpus “Question 1. I am frustrated even a little” from the speaker 6 via the voice output module 41. Then, the output net NET5 is activated and the input net NET4 is deactivated.

(Scenario step 6)
Next, when NET5 is activated, the selection execution unit 43 selects a process item record JA3 having NET5 as an input net. Since this process item record JA3 belongs to the connection point record group, the connection point / end point processing unit 44 activates the output net NET6 of JA3 and deactivates the input net NET5.

(Scenario step 7)
Next, when NET6 is activated, the recognition information input unit 42 selects QB1 having NET6 as an input net. However, since the answer to the question has not yet been input by the examinee, the recognition information input unit 42 continues to cyclically select each processing item record in the processing item table without doing anything.

(Scenario step 8-1)
Next, the voice of the answer to the question is input from the microphone 7 by the examinee. For example, it is assumed that a voice “applicable” is input. When the recognition information input unit 42 selects QB1 with NET6 as an input net and determines that an answer voice is input, the recognition information input unit 42 examines a corpus that matches the answer and selects an output net according to the matching corpus. . In this case, the output net corresponding to the corpus “applicable” is NET7. Accordingly, the answer determination unit 36 activates the selected output net NET7 of QB1 and deactivates the input net NET6.

  When NET7 is activated, the selection execution unit 43 selects a process item record OUT1 having NET7 as an input net. Since this processing item record OUT1 belongs to the answer output record group, the answer output unit 38 stores the answer number of the answer “applicable” to the question 1 in the inquiry result table. Then, the output net NET14 of OUT1 is activated and the input net NET7 is deactivated.

  In the case where the answer voice is “not applicable” or “cannot be said”, the operation similarly follows the scenario of FIG.

(Scenario step 8-2)
If the voice of the answer to the question from the microphone 7 by the examinee is “return to the front”, the answer determination unit 36 selects NET11 as the output net, activates the selected output net NET11 of QB1, The input net NET6 is deactivated. When NET11 is activated, the selection execution unit 43 selects a process item record JA2 having NET11 as an input net. Since this process item record JA2 belongs to the connection point record group, the connection point / end point processing unit 44 activates the output net NET6 of JA2 and deactivates the input net NET11. In this way, the process returns to scenario step 5.

(Scenario step 8-3)
When the voice of the answer to the question from the microphone 7 is “from the beginning” by the examinee, the answer determination unit 36 selects NET12 as the output net, activates the selected output net NET12 of QB1, and inputs Deactivate net NET6. When the NET 12 is activated, the selection execution unit 43 selects the process item record JA6 having the NET 12 as an input net. Since this process item record JA6 belongs to the connection point record group, the connection point / end point processing unit 44 activates the output net NET19 of JA6 and deactivates the input net NET12. Further, when the NET 19 is activated, the selection execution unit 43 selects the process item record JA1 having the NET 19 as an input net. Since this processing item record JA1 belongs to the connection point record group, the connection point / end point processing unit 44 activates the output net NET2 of JA1 and deactivates the input net NET19. Thus, it returns to scenario step 3.

(Scenario step 8-4)
When the voice of the answer to the question from the microphone 7 is “finished” by the examinee, the answer determination unit 36 selects the NET 213 as the output net, activates the selected output net NET 213 of QB1, and Deactivate NET6. When the NET 213 is activated, the selection execution unit 43 selects the process item record END3 having the NET 213 as an input net. Since this process item record END3 belongs to the end point record group, the connection point / end point processing unit 44 deactivates the input net NET 213 and ends the inquiry dialogue process.

(Scenario step 8-5)
When the voice of the answer to the question from the microphone 7 by the examinee does not correspond to any of the corpus of QB1, the answer determination unit 36 selects NET10 as the output net and activates the selected output net NET10 of QB1. The input net NET6 is deactivated. When NET10 is activated, the selection execution unit 43 selects a process item record TTS3 that uses NET10 as an input net. Since this processing item record TTS3 belongs to the guide information record group, the guide sentence output unit 40, via the voice output module 41, the corpus of TTS3 “Answer to the question is“ applicable ”,“ not applicable ”, and“ not applicable ” "Please answer with either of the following." Is output from the speaker 6. Then, the output net NET13 is activated and the input net NET10 is deactivated. Next, when the NET 13 is activated, the selection execution unit 43 selects the process item record JA3 having the NET 13 as an input net. Since this process item record JA3 belongs to the connection point record group, the connection point / end point processing unit 44 activates the output net NET6 of JA3 and deactivates the input net NET13. Thus, the process returns to scenario step 7.

  In this way, the recipient's answer to the question is the expected answer to the question (in this case “applicable”, “not applicable”, “neither”, “return to previous”, “from the beginning” If the answer is an unexpected answer that is neither “End”), a guide sentence that guides the user to make a correct selection answer to the question sentence is output as a voice, and a reply is awaited. Therefore, it is possible to correctly guide the receiver to make an assumed answer.

  In addition, since scenario step 8 and subsequent steps are performed in the same manner in accordance with the interview dialogue scenario of FIG. 13, the subsequent description is omitted.

It is a figure showing the structure of the automatic inquiry system which concerns on Example 1 of this invention. It is a figure showing the functional structure of the automatic inquiry system of FIG. It is a figure showing the data structure of the question information record group preserve | saved at the question information storage part 11. FIG. It is a figure showing the relationship between the data structure of the inquiry result table and the examinee table stored in the inquiry result storage unit 12, and other tables. It is a flowchart showing operation | movement of the automatic inquiry system 1 which concerns on Example 1 of this invention. It is an input screen for the name and the examinee ID displayed on the screen of the touch panel 8. It is a selection screen for an inquiry form displayed on the screen of the touch panel 8. It is a question guide screen displayed on the screen of the touch panel 8. It is a question guide screen displayed on the screen of the touch panel 8. It is an end screen displayed on the screen of the touch panel 8. It is a figure showing the functional structure of the automatic inquiry system which concerns on the present Example 2. FIG. It is a part of transition diagram of the interview dialogue scenario of a stress check questionnaire, a health survey questionnaire, an HDA scale questionnaire, and a TAS-20 questionnaire. It is a transition figure of the whole interview dialog scenario of a stress check questionnaire. It is a figure showing the process item table stored in the process item table memory | storage part. 4 is a diagram illustrating an example of a process item table stored in a process item table storage unit 33. FIG. 4 is a diagram illustrating an example of a net state table stored in a net state table storage unit 34. FIG. 4 is a flowchart showing operations of a voice recognition unit 32 and a recognition information input unit 42. 5 is a flowchart showing the operation of a selection execution unit 43. 10 is a flowchart showing the operation of the health management support system described in Patent Document 2.

Explanation of symbols

Automatic inquiry system 1
Management server 2
Diagnostic computer 3
Patient side terminal device 4
Network 5
Speaker 6
Mike 7
Touch panel 8
Question information storage unit 11
Interview result storage unit 12
Communication interface 13
Audio output unit 14
Voice recognition unit 15
Question group selector 16
Question reading unit 17
Answer determination unit 18
Answer storage unit 19
Interrupt control unit 20
Guide information storage unit 21
Guide output unit 22
Communication interface 23
Computer 30
Recognition dictionary storage unit 31
Voice recognition unit 32
Process item table storage unit 33
Net state table storage unit 34
Scenario starter 35
Answer determination unit 36
Question sentence output part 37
Answer output unit 38
Interview result storage module 39
Guide sentence output unit 40
Audio output module 41
Recognition information input unit 42
Selection execution unit 43
Connection point / end point processing unit 44
Screen display module 45
Input recognition module 46

Claims (4)

An automatic interview system that automatically conducts interviews with patients,
A microphone for the examinee to input voice,
A speaker that outputs sound to the examinee;
A voice recognition means for converting the voice inputted by the examinee from the microphone into character data and outputting it;
Voice output means for converting character data into voice and outputting from the speaker;
Stores the question information, which is a question sentence for the question from the automated question system in the question dialogue dialogue scenario, which is the progress of the dialogue between the automated question system and the examinee in the interview, asking for answers from multiple options. Question information storage means,
An inquiry result storage means for storing answer information, which is information of a selected answer selected by the examinee for the question sentence, in association with the question sentence;
Question reading means for sequentially reading the question information stored in the question information storage means, and outputting the question sentence from the speaker by the voice output means;
Immediately after the question reading means utters and outputs the question sentence, the answer information that the examinee inputs from the microphone and is converted into character information by the voice recognition means is assumed for the immediately preceding question sentence. Answer discriminating means for discriminating whether the answer is an expected answer or other unexpected answer;
Guide information storage means for storing guide information that is guide sentence information for guiding to make a correct selection answer to the question sentence according to the type of each question sentence choice;
When the answer determination means determines that the answer information is the assumed answer, the answer storage means for storing the answer information in association with the question text in the inquiry result storage means;
When the answer determination means determines that the answer information is the unexpected answer, the guide information stored in the guide information storage means is read according to the type of option of the question sentence, and the guide information is An automatic inquiry system comprising: guide output means for outputting an utterance from the speaker by the voice output means. Immediately after the voice output unit utters and outputs the question information, the answer determination unit converts the answer information input from the microphone and converted into character information by the voice recognition unit into the previous question information. It is to determine whether the answer is an assumed answer for selecting the option specified for the interrupt, an interrupt control instruction, or any other unexpected answer.
The interrupt control means for executing an interrupt operation specified by the interrupt control instruction when the answer determination means determines that the answer information is the interrupt control instruction. The automatic inquiry system according to 1. An automatic interview system that automatically conducts interviews with patients,
A microphone for the examinee to input voice,
A speaker that outputs sound to the examinee;
A voice recognition means for converting the voice inputted by the examinee from the microphone into character data and outputting it;
Voice output means for converting character data into voice and outputting from the speaker;
Each information processing unit on the system side in the interview dialogue scenario, which is the progress of the dialogue between the automatic interview system and the examinee in the interview, is called a processing item, and each processing item that accompanies the progress of the dialog process according to the interview dialog scenario When the transition relationship between them is called a net,
A record is assigned to each of the processing items, and each record specifies input net information for specifying an input net on the transition side to the processing item and an output net on the side of transition from the processing item. A processing item table storage means for storing a processing item table storing a corpus that describes output net information and utterance contents or receiving keywords set as necessary for the processing item;
A record is assigned to each of the nets, and each record has a net state table storage for storing a net state table in which the net ID of the net and a net state flag indicating the active state of the net are stored. Means,
An interrogation result storage means for memorizing interrogation results;
With
The processing item table is at least:
As the corpus, a start keyword for starting the interview dialogue scenario is stored, and as the output net information, a net of a net that transitions to a processing item that performs an utterance process of the first question sentence or guide sentence in the interview dialogue scenario A start information record group in which an ID is stored;
As the corpus, a question sentence that is a question sentence for the examinee and that asks for answers by a plurality of options is stored, and the input net information is a network that is connected to a processing item that is a transition source to the question sentence in the inquiry dialogue scenario. A group of question information records in which the net ID of the net connected to the processing item for determining the answer of the examinee to the question sentence in the inquiry dialogue scenario is stored as the output net information;
As the input net information, a net ID of the input net connected to any of the question sentence processing items is stored, and as the corpus, an answer that is a keyword of a selected answer selected by the examinee for the question sentence A plurality of keywords are stored and, as the output net information, a net ID of the output net selected according to the answer keyword and a net ID of the output net selected for an unexpected answer are stored Information records,
As the corpus, a guide sentence and other guide sentences for guiding a correct selection answer to the question sentence corresponding to each question sentence and other guide sentences are stored, and the guide information in the inquiry dialogue scenario is used as the input net information. The net ID of the net connected to the process item that is the transition source to the sentence is stored, and the net ID of the net that is connected to the process item that is the transition destination from the guide sentence in the inquiry dialogue scenario is stored as the output net information. Guide information record group,
Including
When the character data is input from the voice recognition unit, the corpus that matches or is included in the character data among the records of the start information record group is searched, and the output net of the record extracted by the search Scenario start means for obtaining a net ID and activating a net state flag corresponding to the net ID obtained in the net state table storage means;
If all the records of the question information record group are sequentially selected and the net status flag of the net status table corresponding to the input net of the selected record is active, the question text of the selected record is A question that is output by the output means, deactivates the net status flag corresponding to the input net of the selected record of the net status table, and activates the net status flag of the output net of the selected record Sentence output means;
When the character data is input from the voice recognition means, the record of the net status flag of the net status table for the input net is searched for among the records of the answer information record group, and the search When a result record is extracted, it is determined whether any one of the answer keywords of the record matches or is included in the character data. If there is the answer keyword that matches or is included in the character data, The net ID for the answer keyword is acquired, the net status flag corresponding to the net ID acquired in the net status table storage means is activated, the net status flag for the input net is deactivated, and the answer While storing keyword selection results in the inquiry result storage means, If there is no answer keyword that matches or is included in the written data, the net ID for the unexpected answer is acquired, and the net status flag corresponding to the net ID acquired in the net status table storage means is activated. And answer determination means for deactivating the net status flag for the input net;
If all the records of the guide information record group are sequentially selected and the net status flag of the net status table corresponding to the input net of the selected record is active, the guide sentence of the selected record is converted to the voice. A guide for outputting by the output means, deactivating the net status flag corresponding to the input net of the selected record of the net status table, and activating the net status flag of the output net of the selected record Sentence output means;
Automatic interview system characterized by having In the answer information record group, as the input net information, a net ID of the input net connected to any of the question sentence processing items is stored, and the corpus includes a plurality of answer keywords and interrupt control. An interrupt control keyword for instructing is stored, and the output net information is selected for the net ID and unexpected response of the output net selected according to the answer keyword and the interrupt control keyword. Net ID of the output net is stored,
The answer discriminating means is a record in which the net status flag of the net status table for the input net is active in each record of the answer information record group when the character data is inputted from the voice recognition means. When a search result record is extracted, it is determined whether or not any of the answer keywords in the record matches or is included in the character data,
If there is an answer keyword that matches or is included in the character data, the net ID corresponding to the answer keyword or the interrupt control keyword is obtained, and the net corresponding to the net ID obtained in the net state table storage means Activate the status flag, deactivate the net status flag for the input net, and store the selection result of the answer keyword in the inquiry result storage means,
If there is the interrupt control keyword that matches or is included in the character data, the net ID for the interrupt control keyword is acquired, and the net status flag corresponding to the net ID acquired in the net status table storage means And deactivating the net status flag for the input net,
If there is no answer keyword that matches or is included in the character data, the net ID for the unexpected answer is acquired, and the net state flag corresponding to the net ID acquired in the net state table storage means is activated. 4. The automatic inquiry system according to claim 3, wherein the net status flag for the input net is deactivated.

Images