Questionnaire system and method

Abstract

To provide a questionnaire system and method capable of correcting the scores of unreliable answers without manual intervention.SOLUTION: A questionnaire system 1 and a questionnaire method include: questionnaire providing means 51 for providing questionnaire questions; answer acquisition means 51 for acquiring answer data to the questions; answer analysis means 52 for converting the answer data into text data as necessary and storing a score assigned by text analysis in a storage device; additional question execution means 53 for providing additional questions to improve the accuracy of answers to the questions; and answer correction means 54 for converting answer data to the additional questions into text data as necessary and correcting the answers to the questions using the score assigned by text analysis.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present invention relates to a survey system and method that has a follow-up question function for improving the accuracy of evaluation of answers to survey questions. [Background technology]

[0002] Traditionally, customer surveys have been conducted to improve products or services, and the analysis results have been reflected in the development of the products or services. In this type of survey, respondents' ratings for the same questions are compiled and quantitatively evaluated. However, there is a problem that the responses may contain noise that makes them inappropriate ratings, reducing the reliability of the survey.

[0003] Regarding a method for evaluating a questionnaire, for example, Patent Document 1 discloses a technology in which a pattern representing the psychological tendencies of the respondent is calculated based on the responses to a group of questions and a group of dummy questions obtained from the respondent, a coefficient is calculated, and the response results are corrected and presented using this, thereby presenting objective questionnaire results that reflect the psychological tendencies of the respondent. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Laid-Open No. 2004-287736 Summary of the Invention [Problem to be solved by the invention]

[0005] By conducting a survey using a computer, it is possible to collect survey responses at low cost. However, there is a problem that the survey responses may contain noise that makes them inappropriate ratings, reducing the reliability of the survey. In contrast, in face-to-face interviews, the interviewer can ask additional questions to improve the accuracy of the survey, but this comes at a cost.

[0006] An object of the present invention is to provide a questionnaire system and method that can correct the scores of unreliable answers without manual intervention. [Means for solving the problem]

[0007] The present invention comprises the following technical means. [1] A questionnaire system comprising: a questionnaire providing means for providing questionnaire questions; an answer obtaining means for obtaining answer data to the questions; an answer analysis means for converting the answer data into text data as necessary and storing a score assigned by text analysis in a storage device; an additional question execution means for providing additional questions to improve the accuracy of the answers to the questions; and an answer correction means for converting the answer data to text data as necessary and correcting the answers to the questions using the score assigned by text analysis. [2] The questions provided by the questionnaire providing means include quantitative evaluation questions and qualitative evaluation questions, the response analysis means has a quantitative evaluation means that analyzes the quantitative evaluation questions and assigns a score, a qualitative evaluation means that analyzes the qualitative evaluation questions and assigns a score, and a validity evaluation means that evaluates the validity of the score by the quantitative evaluation means based on the score by the qualitative evaluation means, and the follow-up question execution means dynamically generates and provides the follow-up questions based on the evaluation results of the validity evaluation means. [3] The questionnaire system according to [2], wherein the validity evaluation means has a function of determining whether there is a deviation exceeding a threshold between the rating by the quantitative evaluation means and the rating by the qualitative evaluation means. [4] The questionnaire system described in [3], wherein the additional question execution means provides the additional question, which is a qualitative evaluation question generated based on the results of text analysis of the answer to the corrected evaluation question, when it is determined that there is a discrepancy exceeding a threshold between the score obtained by analyzing the answer to the qualitative evaluation question and the score to the quantitative evaluation question. [5] The questionnaire system described in [4], wherein the answer correction means has a function of excluding from the population of the questionnaire all answers relating to the score or answers to the questionnaire including the score when it is determined that the deviation between the score assigned to the additional question and the score assigned to the quantitative evaluation question by the answer analysis means exceeds a threshold value. [6] The quantitative evaluation questions include a first quantitative evaluation question and a second quantitative evaluation question; The qualitative evaluation questions include a first qualitative evaluation question that asks about the reason for the rating of the first quantitative evaluation question, and a second qualitative evaluation question that asks about the reason for the rating of the second quantitative evaluation question. [7] The questionnaire system described in [2], wherein the answer analysis means includes means for calculating an average score of the answers to the quantitative evaluation questions, and means for calculating an average score of the answers to the qualitative evaluation questions based on the average score of the answers to the quantitative evaluation questions. [8] The questionnaire system described in [2], wherein the questionnaire providing means has a function of providing at least a portion of the questionnaire questions to an external information terminal by voice, the answer acquiring means has a function of acquiring at least a portion of the questionnaire answers from the external information terminal by voice data and converting the acquired voice data into text data, and the follow-up question executing means has a function of providing the follow-up questions to the external information terminal. [9] The questionnaire system according to [8], wherein the response analysis means estimates the respondent's emotions by analyzing the voice data, and evaluates the validity of the rating by the quantitative evaluation means in combination with the evaluation by the text analysis.

[10] A survey method in which the following steps are carried out by a computer: a survey execution step of providing survey questions; an answer acquisition step of obtaining answer data to the questions; an answer analysis step of converting the answer data into text data as needed and evaluating it by text analysis; an additional question step of providing additional questions to improve the accuracy of the answers to the questions; and an answer correction step of converting the answers to the additional questions into text data as needed and correcting the answers to the questions using scores assigned by text analysis. [Effects of the Invention]

[0008] According to the present invention, it is possible to provide a questionnaire system and method that can correct the scores of unreliable answers without manual intervention. [Brief explanation of the drawings]

[0009] [Figure 1] 1 is a configuration diagram of an information providing system according to a first embodiment. [Figure 2] 1 is a block diagram illustrating an information providing device according to a first embodiment. [Figure 3] FIG. 10 is a diagram showing an example of questions in a customer survey for the inn / hotel industry. [Figure 4] FIG. 2 is a block diagram illustrating a response analysis means according to the first embodiment. [Figure 5] 10 is a graph output by a response output means according to the first embodiment. [Figure 6] 10 is a flowchart showing a procedure for conducting a survey according to the first embodiment. [Figure 7] 10 is a flowchart showing a procedure for conducting a survey according to the second embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0010] A detailed example of the present invention will be described below with reference to an embodiment. First Embodiment The information provision system 1 of the first embodiment is a questionnaire system that uses artificial intelligence (AI) to analyze the scores of answers to questions given by voice and provides follow-up questions to improve the accuracy of answers to questions that lack objectivity. The following describes the configuration of the information provision system 1 of the first embodiment, followed by a description of its operation.

[0011] [composition] Fig. 1 is a configuration diagram of an information provision system according to this embodiment. As shown in Fig. 1, the information provision system 1 according to this embodiment is composed of an information provision device 10 functioning as a server, a management information terminal 20 functioning as a management client, and an information terminal 30 used by survey respondents, and is configured to be able to communicate with each other via a telecommunications line 40. Note that, although the example shown in Fig. 1 illustrates one information provision device 10 and one management information terminal 20, the configuration is not limited to this, and two or more of each may be used.

[0012] The information providing device 10 is, for example, a PC server, and has a calculation unit 11, a memory unit 12 that stores a dialogue program 50 and a database 60 (described later), and a communication unit 13 that enables communication with the information terminal 30. The information providing device 10 provides a questionnaire to the information terminal 30 used by the questionnaire respondent via a telecommunications line 40.

[0013] The management information terminal 20 is an information terminal for operating the database 60 and executing a management program 70 (described later), and is, for example, a personal computer. The management program 70 includes a question data input means 71 for inputting survey questions into the question DB 61 of the database 60, and a response data inquiry means 72 for outputting the results of an analysis of the survey responses stored in the response DB 62 to an external display device.

[0014] The information terminal 30 is an information terminal with a calling function used by the survey respondents, and is used to transmit response data to the information providing device 10. In the example of FIG. 1, the information terminal 30 is a smartphone, but is not limited to this and can be configured as any information terminal with a calling function, such as a personal computer or tablet. In this embodiment, the survey is provided via a web page, so the information terminal 30 needs to be equipped with a web browser. However, instead, a survey response application may be installed on the information terminal 30.

[0015] 2, the calculation unit 11 of the information providing device 10 executes the dialogue program 50 stored in the storage unit 12, thereby realizing (A) a questionnaire execution means 51, (B) a response analysis means 52, (C) a follow-up question execution means 53, (D) a response correction means 54, and (E) a response output means 55. Each of these means will be described below.

[0016] (A) Survey execution means 51 The survey execution means is a means for providing a survey to the information terminal 30, and includes a survey provision means and a response acquisition means. Various surveys (for example, FIG. 3) are stored in the question DB 61, and any survey can be selected and provided. In this embodiment, when the information terminal 30 reads a QR code (registered trademark, omitted below) and accesses the information provision device 10, the survey is provided to the respondent. The survey may be provided either as audio or as text information displayed on a web page.

[0017] Respondents answer the questionnaire questions by voice or by inputting text. Here, it is preferable to answer by voice, at least for questions that require an evaluation. When answers are requested by voice, answers can be obtained without the revision that would be required for written answers, making it easier to obtain a more straightforward evaluation, and by asking follow-up questions (described below), it is possible to collect a wide range of opinions, as in a focus group interview. Questions regarding attribute information may also be answered by inputting text. Survey response data is stored in the response DB 62. The survey provided by the survey execution means 51 includes at least questions that require answers using a predetermined multi-level rating and questions that require free responses using comments.

[0018] FIG. 3 is a diagram showing an example of questions in a customer survey for the hotel / ryokan industry (accommodation industry). Q1-2 are questions asking for attribute information (hereinafter referred to as "attribute questions"), Q3-6 are questions for quantitative evaluation (hereinafter referred to as "quantitative evaluation questions"), and Q7 is a question for qualitative evaluation (hereinafter referred to as "qualitative evaluation question").

[0019] Q1 is a question asking about the respondent's profile. Age may be answered by indicating the generation. Note that although the questionnaire in Figure 3 is anonymous, the technical concept of the present invention can also be applied to anonymous questionnaires. Q2 asks about travel style. The number of times traveled and the number of people traveling can be multiple choice. Q3 asks about the level of recommendation, asking respondents to answer on a multi-level scale, "How likely is it that you would recommend our hotel to your family and friends?"

[0020] Q4 asks for an overall evaluation and requires a multi-level evaluation. Q5 is a question asking for individual evaluation, and each question item is asked to be answered on a multi-level scale. Q6 asks about the intention to visit again and asks for a multi-level rating.

[0021] Q7 is a question for qualitative evaluation, and requests a free response to the question "Please tell us your opinions and requests." The response to Q7 is analyzed by the qualitative evaluation means 523 described below and is used as material for verifying the validity of the scores for Q3 to Q6. Note that while only one free response is provided in FIG. 3, a free response asking the reason for the score may be requested for each quantitative evaluation question, and the validity may be verified by the qualitative evaluation means 523 based on the free response for each quantitative evaluation question.

[0022] (B) Answer analysis means 52 As shown in FIG. 4, the response analysis means 52 includes a text data conversion means 521, a quantitative evaluation means 522, a qualitative evaluation means 523, and a validity evaluation means 524. The text data conversion means 521 converts the voice data of the answers to the questionnaire received from the information terminal 30 into text data in real time, and stores the text data in the storage unit 12. The quantitative evaluation means 522 analyzes the text data of the answers to the quantitative evaluation questions (Q3 to 6) by natural language processing, assigns scores, and stores the scores in the answer DB 62 as numerical data.

[0023] The qualitative evaluation means 523 uses an AI tool to analyze the text data of the answers to the qualitative evaluation question (Q7), assigns scores to the quantitative evaluation questions (Q3-6) using the same scale as the quantitative evaluation means 522, and stores the scores in the answer DB 62. In a preferred embodiment, the qualitative evaluation means 523 includes sentiment analysis using natural language processing (NLP), extracting and analyzing expressions related to evaluations and emotions contained in the answer text to determine whether the answer is positive or negative and assign a score. The qualitative evaluation means 523 is an AI tool that performs natural language processing using a machine learning model or a deep learning model. In this embodiment, the qualitative evaluation means 523 is implemented using an AI tool (e.g., ChatGPT) that uses large language models (LLMs) constructed using deep learning. Verification using actual survey results confirmed a positive correlation between qualitative evaluation using an AI tool and quantitative evaluation using statistical software. When performing evaluation using an AI tool, background information such as the vendor, customer unit price, season, and location may be entered before the evaluation is performed.

[0024] Furthermore, the qualitative evaluation means 523 has a function of calculating an average score of the answers when there are multiple qualitative evaluation questions. Furthermore, the qualitative evaluation means 523 may be configured to have a function of calculating an average score of the answers to the qualitative evaluation questions based on the average score calculated by the quantitative evaluation means 522. Here, the qualitative evaluation means 523 may have a function of assigning scores so that the average score of the answers to the qualitative evaluation questions is the same as or close to the average score calculated by the quantitative evaluation means 522, or may have a function of assigning scores so that the average score of the answers to the qualitative evaluation questions is a value obtained by adding or subtracting a certain value from the average score calculated by the quantitative evaluation means 522. As an example of the former, for example, when the average score calculated by the quantitative evaluation means 522 is 7 points, the average score is disclosed as being 7 points or in a range of 7 points ± 1 point. As an example of the latter, for example, when the average score calculated by the quantitative evaluation means 522 is 9 points, a score is assigned to the qualitative evaluation question (Q7) so that the average score is 6 points or in the range of 6 points ± 1 point. Initially, the average score of the answers to the qualitative evaluation questions is calculated without reference to the average score calculated by the quantitative evaluation means 522, and when the deviation from the average score calculated by the quantitative evaluation means 522 exceeds a certain level, a function may be provided to recalculate (re-rate) the average score of the answers to the qualitative evaluation questions based on the average score calculated by the quantitative evaluation means 522.

[0025] The validity evaluation means 524 verifies the validity of the scores for the quantitative evaluation questions (Q3 to 6). Specifically, if there is a difference between the score by the qualitative evaluation means 523 and the score by the quantitative evaluation means 522 that is equal to or greater than a threshold, the validity is determined to be low, and the validity evaluation means 524 sends a command to the follow-up question execution means 53 to execute a follow-up question. The validation of the validity by the validity evaluation means 524 in this embodiment is an absolute evaluation performed based only on the answer results of one respondent. In other words, the validation of the validity of the scores for the quantitative evaluation questions (Q3 to 6) is to determine whether the scores for the quantitative evaluation questions are consistent with the answers to the qualitative evaluation questions within the range of the answer results of one respondent.

[0026] An example of validity verification by the response analysis means 52 will be described below. (Verification example 1) In response to Question 1, "What was your overall satisfaction rating? Please also include your reasons," the guest answers, "My overall satisfaction rating is 10 points. The front desk staff didn't greet me, and the wait time before check-in was long." The quantitative evaluation means 522 evaluates this response as 10 points by text analysis, and the qualitative evaluation means 523 evaluates the overall satisfaction rating as 4 points based on the content of the free-form response. If the threshold for Question 1 is set to 3 points, the validity evaluation means 524 determines that an additional question is needed to request a review of the rating for Question 1 because the difference between the rating by the quantitative evaluation means 522 and the rating by the qualitative evaluation means 523 is 6 points, exceeding the threshold, and causes the additional question execution means 53 to provide the additional question. For example, the additional question execution means 53 generates an additional question, such as "Thank you for your high rating of 10 points. How long was the wait time before check-in?" in real time based on the results of natural language analysis of the response, and provides the additional question.

[0027] (Verification example 2) In response to Question 2, "How many points did you rate the facility's facility cleanliness? Please also explain why," the guest answers, "I'd rate the facility's facility cleanliness as a 3. The facilities in the large bath and rooms were well-maintained and felt clean." In this case, the quantitative evaluation means 522 analyzes the text of the response and rates the facility's facility cleanliness as a 3, while the qualitative evaluation means 523 rates the facility's facility cleanliness as an 8 based on the content of the free-form response. If the threshold for Question 1 is set to 4, the validity evaluation means 524 determines that an additional question is necessary to request a review of the rating for Question 2 because the difference between the rating by the quantitative evaluation means 522 and the rating by the qualitative evaluation means 523 is 5, which exceeds the threshold, and causes the additional question execution means 53 to provide the additional question. For example, the additional question execution means 53 generates an additional question in real time based on the results of text analysis of the response, such as "The facility cleanliness was rated as a 3. Which facilities did you feel were not clean enough?" and provides the additional question.

[0028] (C) Additional question execution means 53 The follow-up question execution means 53 provides the respondent with a follow-up question that requests a review of the quantitative evaluation question that the answer analysis means 52 has determined to have low accuracy. The follow-up question requests a response using a multi-level rating scale that is the same as that of the target quantitative evaluation question. The follow-up question may be a question that simply requests a review of the rating of the target quantitative evaluation question, or a question that requests a review of the rating after mentioning that the rating of the free-response was high (or low), or a question that requests a review of the rating after quoting positive (or negative) keywords in the free-response. The follow-up question may be asked not only once, but multiple times.

[0029] (D) Answer correction means 54 The answer correction means 54 analyzes the answers to the follow-up questions sent by the follow-up question execution means 53 and corrects the answers given before the follow-up questions were asked. The answer correction means 54 includes a text data conversion means and a quantitative evaluation means. It converts the audio data of the answers to the received follow-up questions into text data, analyzes the text data of the answers to obtain a score, and overwrites and stores the score in a storage device as the score for the target quantitative evaluation question. For example, if the initial answer to Q4 in FIG. 3 is 4 points and the answer to the follow-up question is 3 points, the answer correction means 54 overwrites and stores the 3 points as the score for Q4. Here, if a follow-up question is asked because there is a discrepancy between the scores for the quantitative evaluation and the qualitative evaluation, but the score for the answer to the follow-up question is the same as the score for the answer to the initial question, and the discrepancy between the scores for the quantitative evaluation and the qualitative evaluation is not bridged, the answer may be determined to be abnormal data (e.g., determined to be from someone who is not answering seriously), and the answer or all of the questionnaire responses including the answer may be deleted from the population.

[0030] (E) Answer output means 55 The answer output means 55 can output the results of evaluating the answers to the quantitative evaluation questions using the NPS (Net Promoter Score: registered trademark, omitted below) index. NPS is an index that can measure customer loyalty (the degree of trust and attachment to a product or service) based on the degree of recommendation. For example, responses to the question, "How likely are you to recommend this product or service to a friend or colleague?" are obtained on an 11-point scale from 0 to 10. Depending on the degree of evaluation, customers are classified as promoters (recommenders), passives, or detractors. A score is calculated by subtracting the percentage of detractors from the percentage of promoters. While customer satisfaction is a short-term indicator that measures the degree of satisfaction with a product or service, customer loyalty is said to be a long-term indicator that indicates the strength of the relationship with a company.

[0031] FIG. 5 is a graph created by the answer output means 55 based on the answers to question Q3 (How likely is it that you would recommend our hotel to your family and friends?). The horizontal axis of the graph represents the ratings on a scale of 1 to 10 from the quantitative evaluation means 522, with blank responses displayed on the far right. The vertical axis of the graph represents the ratings on a scale of 1 to 10 from the qualitative evaluation means 523. In FIG. 5, the graph is illustrated using a box-and-whisker plot, with the bottom of the box representing 25%, the horizontal line within the box representing the median (50%), and the top of the box representing 75%. The average value is displayed with an x, and outliers are displayed with a gray circle. As shown by (1) in FIG. 5, it can be seen that at ratings 8 and 9 on the horizontal axis, the ratings from the quantitative evaluation means 522 are high but the ratings from the qualitative evaluation means 523 are low. 5, it can be seen that for scores 2 and 3 on the horizontal axis, the scores given by the quantitative evaluation means 522 are low, but the scores given by the qualitative evaluation means 523 are high. According to the present invention, by asking follow-up questions to the answers shown in (1) and (2) of FIG. 5, it is possible to increase the accuracy of the answer to question Q3.

[0032] [Operation] FIG. 6 is a flowchart showing a procedure for conducting a survey according to the first embodiment. (STEP 601) When a respondent reads the QR code using the information terminal 30, the survey execution means 51 of the information providing device 10 provides the respondent with survey questions. (STEP 602) The respondent answers each question in the questionnaire by voice using the information terminal 30 and transmits the answers to the information providing device 10. Here, the respondent may answer attribute questions that are not subject to evaluation by entering text.

[0033] (STEP 603) The response analysis means 52 of the information providing device 10 analyzes the responses to the received questionnaire questions. Specifically, the text data conversion means 521 of the response analysis means 52 converts the acquired voice data of the questionnaire responses into text data. Subsequently, the quantitative evaluation means 522 and the qualitative evaluation means 523 of the response analysis means 52 evaluate the questionnaire responses through text analysis. The responses to questions Q3 to Q6 related to the quantitative evaluation are analyzed by the quantitative evaluation means 522 to obtain scores, which are stored as numerical data in the response DB 62. The response to question Q7 related to the qualitative evaluation is analyzed by the qualitative evaluation means 523, and scores assigned using the same scale as the quantitative evaluation means 522 are stored as numerical data in the response DB 62.

[0034] (STEP 604) The validity evaluation means 524 of the response analysis means 52 determines whether there is a deviation exceeding a threshold between the score by the quantitative evaluation means 522 and the score by the qualitative evaluation means 523 for each quantitative evaluation question. (STEP 605, STEP 606) If there is a quantitative evaluation question determined in STEP 604 to have a deviation exceeding the threshold, the additional question execution means 53 provides an additional question requesting a review of the rating of the target quantitative evaluation question. If there are multiple quantitative evaluation questions determined to have a deviation, an additional question is provided for each quantitative evaluation question. If there are no quantitative evaluation questions determined to have a deviation in STEP 604, no additional question is provided and the survey ends.

[0035] (STEP 607) The respondent answers the follow-up question by voice using the information terminal 30 and transmits the answer to the information providing device 10. (STEP 608) The answer correction means 54 of the information providing device 10 analyzes the answers to the follow-up questions in the received questionnaire. Specifically, the answer correction means 54 converts the voice data of the answers to the follow-up questions in the received questionnaire into text data and obtains a score by analyzing the text. If there are multiple follow-up questions, the answer correction means 54 obtains a score for all of the follow-up questions. (STEP 609) The answer correction means 54 overwrites the score of the question corresponding to the acquired follow-up question with the score acquired in STEP 608. If there are multiple follow-up questions, the score of questions corresponding to all of the follow-up questions is overwritten with the score acquired in STEP 608. The above is the procedure for conducting the questionnaire according to the embodiment.

[0036] As described above, the information provision system 1 according to this embodiment can improve the accuracy of a survey by correcting the answers to quantitative evaluation questions that AI has determined to be lacking in objectivity based on the answers to additional questions. In face-to-face interviews, the questioner can ask additional questions to improve the accuracy of the survey, but this can be costly. In this regard, the information provision system 1 according to this embodiment can improve the accuracy of the survey unmanned using AI, thereby making it possible to obtain survey results with the same accuracy as manual interviews at low cost. Furthermore, since there is no need to coordinate schedules for conducting the survey, surveys can be conducted in parallel with many respondents, and respondents can respond at any time, which is convenient. It is also possible to request surveys from overseas respondents without worrying about time zone differences.

[0037] Second Embodiment The information provision system 1 according to the second embodiment includes an information provision device 10 functioning as a server, an administration information terminal 20 functioning as an administration client, an information terminal 30 used by the survey respondents, and a telecommunications line 40, and the hardware configuration is the same as that of the first embodiment. The second embodiment differs from the first embodiment in that the response analysis means 52 of the dialogue program 50 executed on the information provision device 10 has a function of analyzing emotions based on voice data, and the follow-up question execution means 53 has a function of generating follow-up questions in real time in accordance with the answers based on the results of the emotion analysis based on the voice data. Here, by analyzing voice data using a deep learning model, it is possible to estimate gender, age, and emotions (calmness, happiness, anger, sadness, fear, etc.) using known AI tools.

[0038] [Operation] FIG. 7 is a flowchart showing a procedure for conducting a survey according to the second embodiment. (STEP 701) When a respondent reads the QR code using the information terminal 30, survey questions are provided by voice from the survey execution means 51 of the information providing device 10. In the second embodiment, a free response is requested for each question, asking for a rating and the reason for the rating. (STEP 702) When the respondent answers one question in the questionnaire by voice using the information terminal 30, the voice data of the answer is automatically transmitted to the information providing device 10. (STEP 703) The response analysis means 52 of the information providing device 10 analyzes the responses to the received questionnaire questions. Specifically, the response analysis means 52 performs emotion analysis based on the voice data and text analysis by converting the voice data into text data, and stores the results of each analysis in the response DB 62.

[0039] (STEP 704, STEP 705) The answer analysis means 52 of the information providing device 10 evaluates the validity of the answer based on the emotion analysis and text analysis of STEP 703. If the evaluated answer is an answer to a follow-up question, the process proceeds to STEP 709; if it is an answer to a prepared question, the process proceeds to STEP 706. (STEP 706) If the answer analysis means 52 finds a uniqueness between the score in the answer and the reason for the score, it proceeds to STEP 707; if no uniqueness is found, it proceeds to STEP 712. Here, uniqueness refers to a case where there is a discrepancy of more than a threshold between the score given by the answerer and the score obtained by text analysis of the reason for the score, or a case where there is an inconsistency between the score given by the answerer and the emotion analysis based on the voice data (for example, a case where the score is high but the estimated emotion is anger, sadness, or fear). The determination of uniqueness can be performed not only for negative scores but also for positive scores.

[0040] (STEP 707) The follow-up question execution means 53 generates follow-up questions in real time that dig deeper into the unique answers based on the validity evaluation by the answer analysis means 52, and provides them to the respondent by voice. Here, a function may be provided to change the wording of the follow-up questions based on the results of the emotion analysis. For example, it is disclosed that if the estimated emotion is anger, the follow-up questions are worded more politely. (STEP 708) When the respondent answers the questionnaire questions by voice using the information terminal 30, the voice data of the answers is automatically transmitted to the information providing device 10. The processes from STEP 702 to STEP 706 are executed within a time range that does not interfere with the dialogue, so the respondent can answer while having a dialogue as if he or she were being interviewed.

[0041] (STEP 709, STEP 710) If the answer to the additional question is found to be peculiar, the answer analysis means 52 judges the answer of the respondent to be abnormal data (determines that the respondent is not answering seriously, etc.) and marks an abnormal flag with the identification number of the respondent. After the respondent has answered all the questions, the answers of the respondent with the abnormal flag may be deleted from the respondent population as they are considered to be unreliable. (STEP 709, STEP 711) If the answer to the follow-up question is not found to be unique, the answer analysis means 52 corrects the score of the answer to the question corresponding to the follow-up question using the score of the follow-up question. In other words, if a follow-up question is provided when it is determined that the score of the answer to the question is unique, the score of the answer to the question that prompted the follow-up question can be corrected without human intervention by correcting the score of the answer that lacks reliability using the score of the follow-up question.

[0042] (STEP 712, STEP 713) The survey execution means 51 provides the remaining questions to the respondent's information terminal 30 until all questions are completed. In the second embodiment, a validity evaluation (STEP 704) is performed after each question is completed, and if an unusual answer is given, an additional question is asked. Note that the number of additional questions is not limited to one, and if unusual answers continue, multiple additional questions may be asked. (STEP 714) When the respondent answers the next question in the questionnaire by voice using the information terminal 30, the voice data of the answer is automatically transmitted to the information providing device 10, and the process proceeds to STEP 703.

[0043] As described above, according to the information provision system 1 of this embodiment, follow-up questions are dynamically generated for answers that the AI ​​determines to be unique, allowing for unmanned interactive interviews based on the content of the answers, making it possible to collect opinions similar to those in focus group interviews at low cost.

[0044] Although the preferred embodiments of the present invention have been described above, the technical scope of the present invention is not limited to the above-described embodiments. Various modifications and improvements can be made to the above-described embodiments, and such modifications and improvements are also included in the technical scope of the present invention.

[0045] For example, the answer analysis means may be provided with a function for estimating the respondent's emotions by analyzing the voice data and combining the estimation with the evaluation by text analysis to assign a score. Specifically, it is disclosed that if a positive emotion is detected, one point is added to the evaluation by text analysis, and if a negative emotion is detected, one point is subtracted from the evaluation by text analysis. Similarly, the answer analysis means may be provided with a function for estimating the gender and / or age by analyzing the voice data, and for creating and analyzing segments of the respondent.

[0046] Furthermore, although the above-described embodiment has been described as an example targeting the accommodation industry, the present invention can be applied to other industries, such as the food and beverage service industry, the entertainment industry, lifestyle-related service industries such as hairdressing and beauty salons and sports gyms, education and technical skills teaching industry, finance industry, insurance industry, retail industry, information and communications industry, goods rental industry, professional service industry (professional services, etc.), advertising industry, real estate industry, transportation industry, etc. Furthermore, the present invention can be applied not only to the service industry but also to manufacturing industries that sell products and equipment. [Explanation of symbols]

[0047] 1. Information provision system 10...Information providing device 20...Administrative information terminal 30...Information terminal 40...Telecommunications lines 50...Dialogue Program 51...Method of conducting the survey 52…Answer analysis means 53... Additional question execution means 54…Answer correction means 60...Database 70...Management Program

Claims

1. a survey provision means for providing survey questions; an answer acquisition means for acquiring answer data of the question; an answer analysis means for converting the answer data into text data as necessary and storing a score assigned by text analysis in a storage device; a follow-up question execution means for providing a follow-up question to improve the accuracy of the answer to the question; and an answer correction means for converting the answer data to the additional question into text data as necessary and correcting the answer to the question using a score assigned by text analysis.

2. the questions provided by the questionnaire providing means include quantitative evaluation questions and qualitative evaluation questions; the answer analysis means includes a quantitative evaluation means for analyzing the quantitative evaluation questions and assigning a score, a qualitative evaluation means for analyzing the qualitative evaluation questions and assigning a score, and a validity evaluation means for evaluating the validity of the score assigned by the quantitative evaluation means based on the score assigned by the qualitative evaluation means; 2. The questionnaire system according to claim 1, wherein the follow-up question execution means dynamically generates and provides the follow-up question based on the evaluation result of the validity evaluation means.

3. 3. The questionnaire system according to claim 2, wherein the validity evaluation means has a function of determining whether there is a deviation exceeding a threshold between the score given by the quantitative evaluation means and the score given by the qualitative evaluation means.

4. The questionnaire system according to claim 3, wherein the additional question execution means provides the additional question, which is a qualitative evaluation question generated based on the results of text analysis of the answer to the corrected evaluation question, when it is determined that there is a discrepancy exceeding a threshold between the score obtained by analyzing the answer to the qualitative evaluation question and the score to the quantitative evaluation question.

5. The questionnaire system according to claim 4, wherein the answer correction means has a function of excluding from the population of the questionnaire all answers relating to the score or all answers to the questionnaire including the score when it is determined that the deviation between the score assigned to the additional question and the score assigned to the quantitative evaluation question by the answer analysis means exceeds a threshold value.

6. the quantitative evaluation questions include a first quantitative evaluation question and a second quantitative evaluation question; The questionnaire system of claim 2 , wherein the qualitative evaluation questions include a first qualitative evaluation question asking about the reason for the rating of the first quantitative evaluation question, and a second qualitative evaluation question asking about the reason for the rating of the second quantitative evaluation question.

7. 3. The questionnaire system according to claim 2, wherein the response analysis means includes: means for calculating an average score of the responses to the quantitative evaluation questions; and means for calculating an average score of the responses to the qualitative evaluation questions based on the average score of the responses to the quantitative evaluation questions.

8. the questionnaire providing means has a function of providing at least a part of the questionnaire questions to an external information terminal by voice; the answer acquisition means has a function of acquiring at least a part of the answers to the questionnaire as voice data from an external information terminal and converting the acquired voice data into text data; 3. The questionnaire system according to claim 2, wherein the follow-up question execution means has a function of providing the follow-up question to an external information terminal.

9. The questionnaire system according to claim 8 , wherein the response analysis means estimates the respondent's emotions by analyzing the voice data, and evaluates the validity of the rating by the quantitative evaluation means in combination with the evaluation by the text analysis.

10. a survey execution step of providing survey questions; an answer acquisition step of acquiring answer data of the question; a response analysis step of converting the response data into text data as needed and evaluating the data through text analysis; a follow-up question step of providing a follow-up question to improve the accuracy of the answer to the question; and an answer correction step of converting the answers to the additional questions into text data as needed, and correcting the answers to the questions using scores assigned by text analysis.

Images