JP2013191158A - Device, method and program for selecting party to ask question - Google Patents

Abstract

PROBLEM TO BE SOLVED: To support even a questioner, who is not skilled in an interview, getting high-quality data from an answerer.SOLUTION: A device for selecting a party to ask a question comprises conversation analysis means 21, topic field identification means 22, interest score calculation means 23, selection means 24 and output means 5. The conversation analysis means 21 analyzes conversation data, discriminates question data and an answerer who is a party to ask a question, and discriminates answer data to the question data and a real answerer. The topic field identification means 22 identifies a topic field of the question data. The interest score calculation means 23 increases an interest score of the answerer regarding the topic field when the answerer who is the party to ask the question is matched with the real answerer, or decreases the interest score of the answerer who is the party to ask the question regarding the topic field and increases the interest score of the real answer regarding the topic field when the answerer who is the party to ask the question is not matched with the real answerer. The selection means 24 acquires the interest score of each answerer regarding a topic field corresponding to the next question and selects an answerer having the highest interest score as a party to ask the next question.

Description

  The present invention is a technique for supporting an interview, and particularly relates to a question partner selection device, a question partner selection method, and a question partner selection program for selecting a question partner.

  In many research fields, including sociology and nursing science, an interview survey is used to extract the problems of the subject.

  There are various methods for interview surveys. In recent years, a method called “semi-structured interview” has been carried out in the field of engineering research in addition to sociology and nursing. This is a technique in which only rough question items are defined, and the order of questions and the depth of questions are flexibly changed according to the flow of conversation with interview respondents (hereinafter referred to as respondents) (Non-Patent Documents). 1).

"Introduction to qualitative research", Uwe Flick, Shunjusha: pp94-121

  The interview method, which flexibly changes the order of questions and the depth of the question according to the flow of conversation with the respondents, is an effective method for listening to the respondents' real intentions. The person who performs the exercise needs to select the question flexibly according to the flow of the story, and the interviewer is required to have a very advanced interview technique. In particular, in such interviews, there is not necessarily one respondent.

  For example, there are cases where a husband and wife in the same household are both present and ask questions. In such a case, even if the questioner asks a question item prepared in advance, the two respondents do not necessarily have to ask the question equally. From the interviews, the questioners deeply focus on the questions in the relevant field toward each respondent, carefully detecting the topic content that each respondent is strongly interested and interested in and easy to answer. By asking questions, you can collect good quality data.

  “Quality data” here refers to data that enables new discoveries that were not clear until now. Specifically, data with abundant variations in experiences, abstract Instead, it means data based on specific cases, data that includes detailed explanations and thoughts about experiences.

  In other words, when there are multiple respondents, it is necessary to determine during the interview which user can obtain more “quality data” by asking the question and select the question partner flexibly. is there. However, the questioner who is not used to the interview asks the respondent who is not interested in the field, and as a result, there is a problem that good quality data (answer) cannot be obtained.

  The present invention has been made in view of the above circumstances, and an object of the present invention is to select a question partner selection device that supports extracting good-quality data from respondents, even for questioners unfamiliar with interviews, It is to provide a question partner selection method and a question partner selection program.

  In order to achieve the above object, the present invention is a question partner selection device for selecting a question partner in an interview, analyzes input conversation data, determines question data and a respondent as a question partner, and The conversation analysis means for determining the answer data and the actual respondent for the question data, the topic area specifying means for specifying the topic area of the question data, and the respondent of the question partner and the actual answerer match, If the respondent's interest score for the topic area is increased and the respondent of the question partner does not match the actual respondent, the interest score for the topic area of the respondent of the question partner is decreased and the actual respondent Interest score calculation means for increasing the interest score for the topic area, and obtain the interest score of each respondent for the topic area corresponding to the next question. A selection means for selecting a high respondents interested score as an interrogation partner next question, and output means for outputting a question opponent selected by the selecting unit, the.

  The present invention is a question partner selection method for selecting a question partner in an interview, wherein the computer analyzes the input conversation data to determine the question data and the respondent as the question partner, and answers to the question data A conversation analysis step for determining data and an actual respondent, a topic area specifying step for specifying a topic area of the question data, and a respondent of the question partner and an actual answerer match, When the interest score for the topic area is increased and the respondent of the question partner does not match the actual respondent, the interest score for the topic area of the respondent of the question partner is decreased and the topic area of the actual respondent is decreased. The interest score calculation step to increase the interest score for each and the interest score of each respondent for the topic area corresponding to the next question Performing most highly respondents interested scores a selection step of selecting as a question partner next question, and an output step of outputting the interrogation opponent selected by the selection step.

  The present invention is a question partner selection program for causing a computer system to function as the question partner selection device.

  According to the present invention, it is possible to provide a question partner selection device, a question partner selection method, and a question partner selection program for assisting even a questioner who is unfamiliar with an interview to extract high-quality data from a respondent. .

It is a block diagram which shows the structure of the question other party selection apparatus which concerns on embodiment of this invention. It is a figure which shows an example of a question item memory | storage part. It is a flowchart explaining operation | movement of this embodiment.

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

  In general, people tend to state many experiences, detailed explanations, and opinions on topics that they are good at or are interested in. Therefore, in the present embodiment, when there are a plurality of respondents (persons who receive questions) in the interview, the amount of answers to each topic during the interview is calculated for each respondent, and each topic is interested according to the balance. -Estimate which respondents are most interested in. And, it is decided which question of the question items prepared in advance to which respondent to ask the question, and presents it to the questioner (person who conducts the interview). Extract useful data (answers).

  FIG. 1 is a configuration diagram showing the configuration of the question partner selection device of the present embodiment.

  The illustrated question partner selection apparatus includes an input unit 1, a processing unit 2, a question item storage unit 3, and an interest score storage unit 4. Further, the processing unit 2 includes a conversation analysis unit 21, a topic field identification unit 22, an interest score calculation unit 23, and a question / answerer selection unit 24.

  The input unit 1 inputs raw conversation data (voice data or text data) during the interview, and outputs it to the conversation analysis unit 21 together with the input time.

  The conversation analysis unit 21 analyzes the data input from the input unit 1 and converts it into a format that can be interpreted. For example, when the data input from the input unit 1 (such as a microphone) is voice data, voice recognition is performed to determine who is the speaker (a questioner or a respondent among a plurality of respondents). ) And the break of the speaker's conversation (utterance). When the speaker of the separated conversation data is a questioner, the conversation data is referred to as “question data”. When the speaker of the separated conversation data is an answerer, the conversation data is referred to as “answer data”.

  In addition, when the speaker is the questioner's question data, the conversation analysis unit 21 specifies the direction in which the spoken voice is directed and determines the answerer (who asked the question) as the questioning partner. Regarding the determination of the question partner, in addition to the case where the conversation analysis unit 21 automatically recognizes the direction of the spoken voice, when the questioner asks a question, the question partner is determined on the GUI or the like by an input device such as a keyboard. It is good also as inputting.

  Note that the data input to the conversation analysis unit 21 via the input unit 1 is not necessarily voice data. For example, when the interview is performed by a text chat via a network, it is assumed that the input conversation data is text data. In this case, since the questioner and the respondent input the text on different terminals, the questioner / respondent can be identified by distinguishing the input terminals. In addition, when interviewing, it is also possible to manually input the stenographer next to the person who is the speaker (questioner / respondent) to the question partner selection device, and the stenographer also manually inputs the question content and answer. It is done.

  As described above, the conversation analysis unit 21 analyzes the input conversation data and discriminates the question data and the answerer of the question partner (hereinafter, “question partner”), as well as the answer data and the actual data for the question data. The respondent (hereinafter “respondent”) is determined. Then, the conversation analysis unit 21 outputs the question data, the question partner, the answer data, and the respondent, and the input time of the question data and the answer data to the topic field specifying unit 22.

  The question item storage unit 3 stores information related to questions to be asked in the interview. FIG. 2 is a diagram illustrating an example of data stored in the question item storage unit 3. In the illustrated question item storage unit 3, a question ID of a question item to be questioned in an interview, a question item, and a topic field (question category) corresponding to the question item are stored in association with each other.

  The question item is the content of the question item that the questioner plans to perform during the interview. The question item may be in a format that can be referred to by the questioner at the time of the question, the question item may be registered as a sentence, a figure or a picture, or in a format in which a plurality of keywords that can recall the question content are listed. There may be.

  In the topic field, a topic field (the topic field to which the question item belongs) including the contents of the corresponding question item is set. When a question item extends over a plurality of topic areas, a plurality of topic areas may be stored for one question item. The topic area registration format may be a keyword, a sentence format, a figure or a picture.

  The question item and the topic field may be registered in advance by the questioner before the interview, but the person who registers in the question item storage unit 3 is not necessarily the questioner. For example, the topic field is a keyword that is grouped according to the degree of similarity of each question item on the system side such as the question partner selection device after the question item is registered by the questioner, and is common to the question items included in each group A method such as tagging as a topic area is also conceivable.

  The topic area specifying unit 22 receives the question data, the question partner, the answer data, the respondent, and the input time of each data input from the conversation analysis unit 21, refers to the question item storage unit 3, Each question item in the question item storage unit 3 is collated. Note that the topic field identification unit 22 may collate the input answer data with each topic field of the question item storage unit 3 to determine which topic field the answer data belongs to.

  The comparison between the question data and each question item is for determining which question item in the question item storage unit 3 the questioner asked in the conversation data during the interview. For example, the topic field identification unit 22 performs morphological analysis of each sentence with all the question items and question data stored in the question item storage unit 3, calculates the similarity, and has the highest similarity. It can be considered that the question item is a question item indicated by the question data. When the topic field specifying unit 22 specifies the question item in the question item storage unit 3 corresponding to the question data, the topic field setting unit 22 determines the topic field set in the question item storage unit 3 corresponding to the question item. Identify as a field.

  It is not always necessary to automatically check the question data and the question item. For example, there is a method in which the questioner or the like selects / inputs the content of the question during the interview on the GUI.

  The topic area identification unit 22 includes the question ID of the identified question item, the corresponding topic area, the question partner received from the conversation analysis unit 21, the respondent, the answer data, and the input time of the answer data. The result is output to the interest score calculation unit 23. Further, the topic field identification unit 22 outputs the question ID of the identified question item to the question / answerer selection unit 24.

  The interest score calculation unit 23 receives the question ID, the topic field, the question partner, the respondent, the answer data, and the input time from the topic field identification unit 22 and calculates the interest score. Then, the interest score calculation unit 23 outputs (stores) the calculated interest score, the respondent (including the questioner), the topic field, and the input time to the interest score storage unit 4. The interest score is an index indicating how much the respondent is interested / interested in the topic field to which each question data belongs. The interest score is calculated, for example, according to the following rule.

  (1) When one respondent responds to one question data, and the question partner and the respondent match (when the question partner and the respondent are the same), the interest of the respondent Increase score. That is, the answer amount of the respondent is calculated as the interest score of the subject field of the respondent and added. This is because the larger the amount of answers, the stronger the interest in the topic area.

  The answer amount is a value calculated based on the answer data. For example, the number of characters or words of the answer data, or the time when the answer data was spoken (speech time) can be used as it is. In addition, since the utterance speed increases when interested in a certain topic field, a value obtained by dividing the number of characters in the answer data by the utterance time (that is, the utterance speed) may be used as the answer amount. Also, a method of calculating the answer amount by the above combination is conceivable (for example, the number of characters in the answer data + the utterance speed). In addition, for example, humans tend to make louder voices and higher voice tones when speaking in areas of special interest or high interest. It is also possible to use a method of calculating by using (for example, adding a score when an input having a magnitude equal to or larger than a certain threshold value is matched).

  In this case, the interest score calculation unit 23 stores the answerer as the question partner, the topic field, the interest score, and the input time as a set in the interest score storage unit 4.

  (2) When one respondent responds to one question data, and the question partner does not match the respondent (when the question partner and the respondent are different), it was originally the question partner It is determined that the person is not interested in the topic area, and the interest score of the topic area of the question partner (respondent as the question partner) is reduced. As a method for reducing the interest score, for example, a negative constant value may be used as the interest score.

  On the other hand, the interest score of the respondent who actually answered is determined to be high in the topic area, and the interest score of the respondent is increased. That is, the answer amount of the respondent is calculated as the interest score of the subject field of the respondent and added. In this case, the fact that the answer has been answered in spite of not being asked by myself indicates that there is a strong interest in the field, so a value higher than the answer amount is calculated as the interest score and added. Also good. As a method for calculating a value higher than the answer amount, any method may be used as long as the interest score is higher than the answer amount, for example, by multiplying the answer amount by 1.5 times or adding a constant to the answer amount.

  In this case, the interest score calculation unit 23 sets the answerer who originally was the question partner, the topic area, the interest score (such as a negative constant value), and the input time as a set and stores them in the interest score storage unit 4. To do. Also, the interest score calculation unit 23 sets the topic area of the respondent who actually responded, the interest score (the answer amount or a value higher than the answer amount), and the input time as a set, and stores the interest score in the interest score storage unit 4. Remember.

  (3) When multiple respondents respond to one question data, and there is a respondent who matches the question partner, the respondent who matches the question partner is interested in the same way as in (1). The score (answer amount) is calculated and stored in the interest score storage unit 4. On the other hand, for each of all the respondents other than the question partner, the interest score (the answer amount or a value higher than the answer amount) is calculated and stored in the interest score storage unit 4 as in (2).

  (4) When multiple respondents respond to one question data, and there is no respondent who matches the question partner, for those who were originally the question partner, an interest score is the same as in (2) (Negative constant value) is calculated and stored in the interest score storage unit 4. For each of all the respondents other than the question partner, the interest score (the answer amount or a value higher than the answer amount) is calculated and stored in the interest score storage unit 4 as in (2).

  (5) Regarding the total response volume (sum of the response volumes of all respondents from the start of the question to the present), the response rate of one respondent is too high or the rate of remarks is high In some cases, it is assumed that the respondent is not interested in the topic area, but simply wants to talk. In this case (that is, when the situation of (5) is applicable and also applies to the situation of (2), (3) or (4)), the interest score calculation method is (2), (3). Or it is possible to change from (4).

  Specifically, even if a respondent who is different from the question partner in (2) answers, the person who was originally the question partner does not determine that he / she is less interested in the topic area, and increases or decreases the interest score. Do not subtract. Also, the interest score of the actual respondent is not determined to be high interest in the topic area, and there is a method such as not increasing / decreasing (increasing) the interest score of the topic area or setting the answer amount as it is as the interest score. Conceivable.

  In addition, even when multiple respondents respond to one question data of (3) and (4), the answer amount (or answer ratio) of the respondent among the respondents other than the question partner For a large number of respondents, a method may be considered in which the interest is not determined to be high in the topic area, and the interest score in the topic area is not increased (increased), or the answer amount is directly used as the interest score.

  The interest score storage unit 4 stores the interest score calculated by the interest score calculation unit 23 for each respondent and each topic area. The interest score storage unit 4 according to the present embodiment is not a method in which the interest score stored in the interest score storage unit 4 is increased or decreased and updated every time the interest score calculation unit 23 calculates an interest score. A data addition method for outputting (storing) data (record) including a question partner), a topic field, an interest score, and an input time to the interest score storage unit 4 is set.

  The question / answerer selection unit 24 refers to the question item storage unit 3 and selects the next question item and its topic area. The following question items may be selected in any way. For example, the question items are selected in the order of the question ID, and the question / answerer selection unit 24 has the next ID of the question ID notified from the topic field specifying unit 22 (question ID of the question item specified from the question data). A method of selecting a question item can be considered.

  When the question / answerer selection unit 24 acquires the next question item and the topic field corresponding to the next question item from the question item storage unit 3, each question corresponding to the topic field from the interest score storage unit 4 using the acquired topic field as a search key. Read the respondent's interest score (answer history). Then, the question / answerer selection unit 24 selects a question partner for the next question item based on the acquired interest score of each answerer. Specifically, among the plurality of respondents, the respondent with the highest interest score is determined as the question partner of the next question item.

  When a plurality of interest scores (answer histories) are stored in the interest score storage unit 4 for one topic field by one respondent, for example, the average value of the plurality of interest scores is calculated as the interest of the topic field. A method is conceivable, in which the maximum value of a plurality of interest scores is used as a score, and the total value of a plurality of interest scores is used as the interest score of the topic field.

  Then, the question / answerer selection unit 24 outputs the selected question partner to the output unit 5 together with the selected question item. The output unit 5 outputs the question partner and the question item to an output device such as a display and presents it to the questioner. The presentation method may be in any form such as voice, image, text, and the like.

  As the question partner selection device described above, for example, a general-purpose computer system including a CPU, a memory, an external storage device such as an HDD, an input device, and an output device can be used. In this computer system, each function of the question partner selection device is realized by the CPU executing a program for the question partner selection device loaded on the memory. Further, the program for the question partner selection apparatus can be stored in a computer-readable recording medium such as a hard disk, a flexible disk, a CD-ROM, an MO, a DVD-ROM, or can be distributed via a network.

  FIG. 3 is a flowchart showing the operation of the question partner selection apparatus of this embodiment.

  The conversation analysis unit 21 waits for the conversation data to be input (S1). When the conversation data is input (S2: present), the topic field identification unit 22 analyzes the conversation data to obtain the questioner and the question data. The respondent, the answer data, and the input time are discriminated and output to the topic field specifying unit 22 (S3). The topic field identification unit 22 collates the question data with each question item in the question item storage unit 3, identifies the topic field of the question item corresponding to the question data, and relates to the topic field, the questioner, the respondent, and the answer The data and the input time are output to the interest score calculation unit 23. Further, the topic field identification unit 22 outputs the question ID of the question item corresponding to the question data to the question / answerer selection unit 24 (S4).

  The interest score calculation unit 23 calculates an interest score based on the data input from the interest score calculation unit 23, and sets the respondent (including the question partner), topic field, interest score, and input time as a set. It memorize | stores in the score memory | storage part 4 (S5). The question / answerer selection unit 24 selects the question item of the question ID next to the question ID of the question item specified in S4 and the corresponding topic field (S6). Then, the question / answerer selection unit 24 acquires the interest score of each respondent in the selected topic area from the interest score storage unit 4, selects the answerer with the highest interest score as the question partner, and selects the selected question partner And the question item selected in S6 are output to the output unit 5 (S7). The output unit 5 outputs the question partner and the question item (S8).

  In the present embodiment described above, in an interview with one questioner and a plurality of respondents, each question belongs depending on whether or not the question partner and the respondent match, and the amount of utterances in the answer data of the respondent. Determine if you are interested in the topic area and, depending on the result, indicate to the respondent who the next question should be addressed to (ie, the person who asked the next question) ,Recommendation to. Thereby, in this embodiment, even if it is a questioner unfamiliar with an interview, it can support extracting good quality data from a respondent. That is, even a questioner who is not accustomed to an interview can ask a question to an appropriate respondent, and can extract high-quality data from the user.

  In addition, this invention is not limited to the said embodiment, Many deformation | transformation are possible within the range of the summary.

1: Input unit 2: Processing unit
21: Conversation analysis unit 22: Topic field identification unit 23: Interest score calculation unit 24: Question / answerer selection unit 25: Face image selection unit 3: Question item storage unit 4: Interest score storage unit 5: Output unit

Claims (8)

A question partner selection device for selecting a question partner in an interview,
Analyzing the input conversation data, determining the question data and the respondent as the question partner, and conversation analysis means for determining the answer data and the actual respondent to the question data;
A topic area specifying means for specifying a topic area of the question data;
If the respondent of the question partner matches the actual respondent, the interest score for the topic area of the respondent increases, and if the respondent of the question partner does not match the actual respondent, the answer of the question partner An interest score calculation means for decreasing an interest score related to the topic area of the person and increasing an interest score related to the topic area of the actual respondent;
A selection means for obtaining an interest score of each respondent regarding a topic area corresponding to the next question, and selecting a respondent with the highest interest score as a question partner of the next question;
And an output means for outputting the question partner selected by the selecting means. The question partner selection device according to claim 1,
The question partner selection device, wherein the interest score calculation means increases the interest score by using an answer amount based on the answer data. The question partner selection device according to claim 1,
The interest score calculation means adds an answer amount based on the answer data to an interest score related to the topic area of the respondent when the respondent of the question partner matches an actual respondent, A question partner selection device characterized by adding a value larger than the amount of answers based on the answer data to the interest score for the topic field of the actual answerer when the answerer and the actual answerer do not match. The question partner selection device according to any one of claims 1 to 3,
A question storage means for storing each question item and a topic area corresponding to the question item,
The topic area specifying means specifies the topic area of the question item corresponding to the question data with reference to the question storage means,
The selection means selects a question item and topic area of the next question from the question storage means,
The said output means outputs the question item of the next question with the question partner selected by the said selection means. The question partner selection apparatus characterized by the above-mentioned. A question partner selection method for selecting a question partner in an interview,
Computer
Analyzing the input conversation data, determining the answer data as the question data and the question partner, a conversation analysis step for determining the answer data for the question data and the actual respondent,
A topic area specifying step for specifying a topic area of the question data;
If the respondent of the question partner matches the actual respondent, the interest score for the topic area of the respondent increases, and if the respondent of the question partner does not match the actual respondent, the answer of the question partner An interest score calculating step of decreasing an interest score related to the topic field of the person and increasing an interest score related to the topic field of the actual respondent;
Selecting each respondent's interest score for the topic area corresponding to the next question, and selecting the respondent with the highest interest score as the next question question;
And a step of outputting the question partner selected in the selection step. The question partner selection method according to claim 5,
The interest score calculating step increases the interest score by using an answer amount based on the answer data. The question partner selection method according to claim 5 or 6,
The computer further includes a question storage unit that stores each question item and a topic field corresponding to the question item,
The topic area identification step identifies a topic area of a question item corresponding to the question data with reference to the question storage unit,
The selection step selects a question item and a topic area of a next question from the question storage unit,
In the output step, the question item of the next question is output together with the question partner selected in the selection step.   A question partner selection program for causing a computer system to function as the question partner selection device according to any one of claims 1 to 4.

Images