Question estimation system, question estimation method, and question estimation program

Abstract

This question estimation system comprises an acquisition unit, a conversion unit, an estimation unit, a replacement unit, and an output unit. The acquisition unit acquires, as an input keyword, a keyword of an assumed question to be generated. The conversion unit converts the input keyword into a conversion keyword that is a synonymous or similar keyword. The estimation unit estimates, on the basis of the conversion keyword, a sentence of the assumed question by using a question estimation model for estimating a sentence of the assumed question from a keyword. The replacement unit replaces, with the input keyword, a conversion keyword included in the sentence of the assumed question estimated by the estimation unit. The output unit outputs the sentence of the assumed question obtained through replacement by the replacement unit.

Description

Technical Field

[0004] , , , , , , , , ,

[0006] , , , ,

[0005] , , , , ,

[0007] , , ,

[0001] The present disclosure relates to a question estimation system and the like.

Background Art

[0002] A conference organizer may, as a preparatory measure, anticipate questions from the attendees and prepare answers to those questions. For example, at an annual general meeting of shareholders, in order to give appropriate answers to questions from the shareholders, the organizer of the annual general meeting of shareholders may, for example, as a preparatory measure, anticipate questions and prepare answers to those questions. Questions are often asked, for example, about current events that are the topic at the time of the conference. Also, the attendees of the conference and the subjects of the questions are diverse, and anticipating questions and preparing answers require a lot of time.

[0003] The search control program of Patent Document 1 searches for an answer to a question using a learning model that has learned the relationship between the keywords included in the question and the keywords included in the answer.

Prior Art Documents

Patent Documents

[0004] [[ID=2⑥]]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] In the search control program of Patent Document 1, it may be difficult to estimate a question according to the keywords.

[0006] An object of the present disclosure is to provide a question estimation system and the like that can easily estimate a question according to the keywords in order to solve the above problems.

Means for Solving the Problems

[0007] To solve the above problems, the question estimation system of this disclosure comprises: an acquisition means for acquiring keywords of the hypothetical questions to be generated as input keywords; a conversion means for converting the input keywords into conversion keywords which are synonymous or similar keywords; an estimation means for estimating the sentence of a hypothetical question using a question estimation model that estimates the sentence of a hypothetical question from input terms based on the conversion keywords; a replacement means for replacing the conversion keywords contained in the sentence of the hypothetical question estimated by the estimation means with the input keywords; and an output means for outputting the sentence of the hypothetical question substituted by the replacement means.

[0008] The question estimation method described herein obtains keywords of the hypothetical questions to be generated as input keywords, converts the input keywords into synonymous or similar keywords called conversion keywords, estimates the sentences of the hypothetical questions using a question estimation model that estimates the sentences of the hypothetical questions from the input terms based on the conversion keywords, replaces the conversion keywords contained in the estimated sentences of the hypothetical questions with the input keywords, and outputs the replaced sentences of the hypothetical questions.

[0009] The recording medium of this disclosure non-temporarily records a question estimation program that causes a computer to execute the following processes: acquiring keywords of the hypothetical questions to be generated as input keywords; converting the input keywords into conversion keywords which are synonymous or similar keywords; estimating the sentence of the hypothetical question using a question estimation model that estimates the sentence of the hypothetical question from the input terms based on the conversion keywords; replacing the conversion keywords contained in the estimated sentence of the hypothetical question with the input keywords; and outputting the substituted sentence of the hypothetical question. [Effects of the Invention]

[0010] According to this disclosure, questions corresponding to keywords can be easily estimated. [Brief explanation of the drawing]

[0011] [Figure 1] This figure shows an overview of the configuration of the disclosed embodiment. [Figure 2]This figure schematically illustrates an example of the operation for estimating anticipated questions in the embodiments of this disclosure. [Figure 3] This figure shows an example of the configuration of the question estimation system according to an embodiment of the present disclosure. [Figure 4] This figure shows an example of a display screen for the estimated results of a question in an embodiment of the present disclosure. [Figure 5] This figure shows an example of a display screen for the estimated results of a question in an embodiment of the present disclosure. [Figure 6] This figure shows an example of a display screen for the estimated results of a question in an embodiment of the present disclosure. [Figure 7] This figure shows an example of the operation flow of the question estimation system in this disclosure. [Figure 8] This figure shows an example of the hardware configuration of the question estimation system in this disclosure. [Modes for carrying out the invention]

[0012] Embodiments of this disclosure will be described in detail with reference to the figures. Figure 1 is a diagram showing an overview of the configuration of the information processing system of this embodiment. The information processing system of this embodiment includes, for example, a question estimation system 10 and a terminal device 20. The question estimation system 10 is connected to the terminal device 20, for example, via a network. There may also be multiple terminal devices 20. The number of terminal devices 20 can be set as appropriate.

[0013] The question estimation system 10 is a system that estimates questions that are expected to be asked by attendees in a meeting. These expected questions are also called anticipated questions. The question estimation system 10 estimates anticipated questions using a question estimation model. The question estimation model is a learning model that estimates questions that are expected to be asked in relation to a given keyword. A question related to a keyword is, for example, a question asked using a question sentence that includes the keyword in the text. Another type of question related to a keyword is, for example, a question in which the keyword is strongly related to the main point of the question.

[0014] The anticipated questions to be considered are, for example, questions asked at a meeting. They could also be questions asked by shareholders to a company at a general shareholders' meeting. They could also be questions asked in a parliament or committee. Furthermore, they could be questions asked outside of meetings. For example, they could be questions asked at a lecture, presentation, exhibition, press conference, media briefing, investor briefing, product launch, product description, service description, or product usage explanation. They could also be questions in a Q&A published in print or on the web. However, the anticipated questions are not limited to those mentioned above.

[0015] The question estimation system 10 converts, for example, a keyword entered by the user into a keyword that is synonymous with the keyword entered by the user and suitable for estimating the anticipated question in the question estimation model. Synonymy may include similar terms. A keyword suitable for estimating the anticipated question in the question estimation model is, for example, a keyword from which the question estimation model can estimate the anticipated question. A keyword suitable for estimating the anticipated question in the question estimation model is, for example, a keyword included in the training data when the question estimation model is generated. The keyword entered by the user to obtain the estimated result of the anticipated question is called, for example, the input keyword. The keyword suitable for input to the estimation model, which is the converted keyword, is called, for example, the converted keyword. The input keyword and the converted keyword are, for example, synonymous or similar terms to each other. Similar terms may include, for example, the converted keyword having a higher-level or lower-level meaning of the input keyword. The user is, for example, the subject who uses the estimated result of the anticipated question to prepare for question assumptions and answers.

[0016] The question estimation system 10, for example, converts an input keyword into a conversion keyword suitable for estimating an assumed question in a question estimation model. Also, the question estimation system 10 estimates an assumed question using the conversion keyword as an input to the question estimation model. Then, the question estimation system 10 replaces the conversion keyword included in the assumed question estimated by the question estimation model with the input keyword. The assumed question replaced from the conversion keyword to the input keyword becomes an assumed question including the input keyword. In this way, the question estimation system 10 estimates an assumed question in which the input keyword is used from the input keyword which is a keyword input by the user.

[0017] FIG. 2 is a diagram schematically showing an example of an operation in which the question estimation system 10 estimates an assumed question. The question estimation system 10, for example, acquires an input keyword from the terminal device 20. The input keyword is input to the terminal device 20, for example, by the user's operation. The question estimation system 10, for example, uses a conversion model to convert the input keyword into a conversion keyword. The conversion model converts the input keyword into a conversion keyword that is synonymous with the input keyword and suitable for estimating an assumed question in the question estimation model. For example, when the input keyword is a term "novel C virus", the question estimation system converts it into a conversion keyword "novel influenza". In this case, "novel C virus" is a term that has, for example, little track of being used in questions and whose frequency of use in news or daily conversations is increasing. And "novel influenza" is a term that has, for example, a lot of track of being used in questions.

[0018] The question estimation system 10 estimates a question from the conversion keyword using the question estimation model. Then, the question estimation system 10 converts the conversion keyword included in the sentence of the estimated question into the input keyword. The question estimation system 10 outputs, as an estimation result of the assumed question, the sentence obtained by converting the conversion keyword included in the sentence of the estimated question into the input keyword. The question estimation system 10 outputs, for example, the estimation result of the assumed question to the terminal device 20.

[0019] Newly coined terms and terms with increasing frequencies are likely to be used as keywords for questions. However, since past question-and-answer performance data is used as learning data when generating a question estimation model, for example, newly coined terms are likely not included in the learning data when generating a question estimation model. Also, terms with increasing frequencies may not be included in the learning data for generating a question estimation model, or even if they are included, the number may not be sufficient. Therefore, when estimating an assumed question using a question estimation model generated using learning data that does not include terms used as keywords, it may not be possible to appropriately estimate assumed questions related to the keywords. On the other hand, in the question estimation system 10, by converting the input keyword into a converted keyword using a conversion model and estimating the question using the converted keyword as the input to the question estimation model, it is highly likely that the question estimation model can perform appropriate estimation. Therefore, the question estimation system 10 can accurately estimate assumed questions related to the keyword even when a keyword not used during the learning of the question estimation system is input. Also, the user does not need to consider, for example, whether the input keyword is suitable as the input to the question estimation model. Therefore, the user can easily obtain candidates for assumed questions related to the keyword.

[0020] Here, a specific example of the configuration of the question estimation system 10 will be described. FIG. 3 is a diagram showing an example of the configuration of the question estimation system 10. The question estimation system 10 basically includes an acquisition unit 11, a conversion unit 12, an estimation unit 13, a replacement unit 14, and an output unit 15. Also, the question estimation system 10 includes, for example, a storage unit 16.

[0021] The acquisition unit 11 acquires keywords of the hypothetical questions to be generated as input keywords. The acquisition unit 11 acquires keywords of the hypothetical questions to be generated as input keywords from, for example, the terminal device 20. The input keywords are entered into the terminal device 20 by the user's operation. The acquisition unit 11 may also acquire sentences that contain the hypothetical questions or parts of the hypothetical questions.

[0022] When multiple conversion models are used in the conversion unit 12, the acquisition unit 11 may acquire the selection of the conversion model to be used for converting the input keyword. Also, when multiple question estimation models are used in the estimation unit 13, the acquisition unit 11 may acquire the selection of the question estimation model to be used for estimating the anticipated question. The selection of the conversion model and the question estimation model is input, for example, to the terminal device 20 by user operation. The acquisition unit 11 then acquires the selection of the conversion model and the question estimation model from the terminal device 20, for example.

[0023] When a conversion keyword to be used for estimating a hypothetical question is selected from the candidate conversion keywords, the acquisition unit 11 may acquire the selection of the conversion keyword to be used for estimating a hypothetical question. The selection of the conversion keyword to be used for estimating a hypothetical question is input, for example, to the terminal device 20 by user operation. The acquisition unit 11 then acquires the selection of the conversion keyword to be used for estimating a hypothetical question from the terminal device 20, for example.

[0024] The acquisition unit 11 may acquire the selection of the field of the anticipated questions. The field of the anticipated questions is, for example, a division of the department, industry, academic field, or organization that the content of the anticipated questions will cover. The field of the anticipated questions is, for example, a division of legal affairs, finance, research, development, manufacturing, sales, transportation, distribution, medical care, education, parliament, public hearings, committees, and management meetings. The field of the anticipated questions may also be, for example, a division of mathematics, physics, chemistry, information science, architecture, civil engineering, literature, art, history, music, and sports. The field of the anticipated questions may also be a more detailed division. The field of the anticipated questions is not limited to the above. The selection of the field of the anticipated questions is input, for example, by the user's operation to the terminal device 20. The acquisition unit 11 then acquires the selection of the field of the anticipated questions from the terminal device 20, for example.

[0025] The conversion unit 12 converts the input keyword into a conversion keyword that is synonymous or similar to the input keyword. For example, the conversion unit 12 converts input keywords that use newly introduced terms or terms whose frequency of use is increasing into a conversion keyword that is synonymous or similar to the input keyword and that the question estimation model can use to estimate the sentence of the hypothetical question. Converting to a conversion keyword that the question estimation model can use to estimate the sentence of the hypothetical question means, for example, converting to a conversion keyword that is already frequently used. Converting to a conversion keyword that is already frequently used means, for example, converting to a term that is frequently included in the training data used to generate the question estimation model. Frequently included in the training data means, for example, that the term is included in the training data to a degree that the estimation accuracy of the hypothetical question generated using the training data is appropriate for the intended use of the hypothetical question.

[0026] The conversion unit 12 may convert one input keyword into multiple conversion keywords. The conversion unit 12 may convert one input keyword into multiple conversion keywords that have the meaning of a sub-concept of the input keyword. Furthermore, if the input keyword has different meanings for each field of the anticipated question, the conversion unit 12 may convert one input keyword into a conversion keyword for each field of the anticipated question. Furthermore, if the input keyword is a coined word formed by combining two or more terms, the conversion unit 12 may convert one input keyword into a conversion keyword corresponding to each term that makes up the coined word.

[0027] When the acquisition unit 11 acquires a document containing a hypothetical question or part of a hypothetical question, the conversion unit 12 converts, for example, the terms contained in the document acquired from the acquisition unit 11 into conversion keywords. The conversion unit 12 extracts, for example, newly used terms and terms whose frequency of use is increasing from the document acquired by the acquisition unit 11 as input keywords. Then, the conversion unit 12 converts the extracted input keywords into conversion keywords.

[0028] The conversion unit 12 converts input keywords to conversion keywords using, for example, a table that defines the relationship between input keywords and conversion keywords. The table defining the relationship between input keywords and conversion keywords may be set up for each field of the anticipated question. The conversion unit 12 converts input keywords to conversion keywords using, for example, a table corresponding to the selection of the field of the anticipated question acquired by the acquisition unit 11. The conversion unit 12 may also convert input keywords to conversion keywords using dictionary data. The conversion unit 12 converts input keywords to conversion keywords using, for example, dictionary data corresponding to the selection of the field of the anticipated question acquired by the acquisition unit 11.

[0029] The conversion unit 12 may convert input keywords to conversion keywords using a conversion model that converts input terms to synonymous or similar terms. The conversion model is, for example, a learning model that converts input terms to synonymous or similar terms. The conversion model takes input keywords as input and converts them to conversion keywords that are synonymous or similar to the input keywords. The conversion model may convert one input keyword to multiple conversion keywords.

[0030] A transformation model is a machine learning model that is generated by learning the relationship between input terms and the transformed terms. It is also generated by learning the relationship between terms and their synonyms or similar terms. For example, a transformation model can be generated by deep learning using a neural network.

[0031] The conversion model is generated using, for example, at least one of the following as training data: a thesaurus, a neologism dictionary, a slang dictionary, and a dictionary of current affairs terminology. The conversion model may also be generated using publicly available terminology data on a network as training data. The conversion model may also be generated using term annotations found in at least one of the following as training data: newspapers, magazines, books, official announcements, and articles. The conversion model may also be generated using a specialized terminology dictionary in the field of the anticipated question as training data. The accuracy of the question estimation model's estimation of anticipated questions may be improved. The training data used to generate the conversion model is not limited to those described above.

[0032] Furthermore, the training data used to generate the transformation model includes, for example, terms included in the training data used to generate the question estimation model. The transformation model is generated by learning the relationship between the terms included in the training data used to generate the question estimation model, for example, as transformed words, and the input terms. The question estimation model is highly likely to fail to accurately predict anticipated questions from newly used terms and terms whose frequency of use is increasing. However, by generating a transformation model using training data that includes terms included in the training data used to generate the question estimation model, and estimating anticipated questions using the transformed keywords obtained using this transformation model, the accuracy of the question prediction model's estimation of anticipated questions may improve. In other words, by generating a transformation model using training data that includes terms included in the training data used to generate the question estimation model, the suitability of the transformed keywords as input keywords to the question estimation model improves. By improving the suitability of the transformed keywords as input keywords to the question estimation model, the accuracy of the question estimation model's estimation of anticipated questions may improve. In addition, the transformation model may be generated by retraining an already generated transformation model using at least one of newly used terms and terms whose frequency of use is increasing. The conversion model is generated, for example, in a system outside the question estimation system 10. The conversion model may also be generated in a generation unit (not shown) provided by the question estimation system 10.

[0033] The conversion unit 12 may convert input keywords to conversion keywords using one of several conversion models. The choice of which of the multiple conversion models to use may be made, for example, by the user. The conversion unit 12 may also select a conversion model to use for conversion from among several conversion models based on a question estimation model used to estimate the anticipated question. The conversion unit 12 may also convert input keywords to conversion keywords using a conversion model corresponding to the field of the anticipated question.

[0034] The conversion unit 12 may convert an input keyword into a conversion keyword using multiple conversion models. That is, the conversion unit 12 may convert a single input keyword into multiple conversion keywords using each of the multiple conversion models.

[0035] The conversion unit 12 may convert input keywords into conversion keywords if the input keywords meet criteria set based on the novelty of the term. For example, if the input keyword is a new term, the question estimation model is unlikely to be able to predict the expected question. On the other hand, if the input keyword is a term that is already in use, the question estimation model is likely to be able to predict the expected question even if the input keyword is entered directly. Therefore, when the input keyword is a new term, the conversion unit 12 converts the input keyword into a conversion keyword and uses it as input to the question estimation model, which allows for efficient estimation of expected questions. For example, the conversion unit 12 converts input keywords that match a term into conversion keywords if the elapsed time since the start of use set for each term is less than the criterion. The criteria for whether or not to convert to conversion keywords are set, for example, by the operator of the question estimation system 10.

[0036] The conversion unit 12 may convert the input keywords into conversion keywords if the question estimation model is unable to estimate the expected question from the input keywords. For example, when the input keywords are used as input to the question estimation model to estimate the expected question, if the input keywords are new terms, the question model is highly likely to be unable to estimate the expected question from the input keywords. For this reason, when the input keywords are used as input to the question estimation model to estimate the expected question, and the model is unable to estimate the expected question from the input keywords, the conversion unit 12 converts the input keywords into conversion keywords. The conversion unit 12 may also modify the conversion model to convert the input keywords into conversion keywords if the question estimation model is unable to estimate the expected question from the conversion keywords.

[0037] Furthermore, the conversion unit 12 may convert the input keywords into conversion keywords if the number of hypothetical questions estimated by the question estimation model from the input keywords is less than a certain threshold. For example, if the number of hypothetical questions estimated by the question estimation model from the conversion keywords is less than a certain threshold, the conversion unit 12 modifies the conversion model and converts the input keywords back into conversion keywords. The threshold for the number of hypothetical questions when converting back into conversion keywords is set, for example, by the operator of the question estimation system 10.

[0038] The conversion unit 12 may determine a priority for each conversion keyword converted from the input keyword to be used in generating hypothetical questions. For example, if the input keyword can be converted into multiple conversion keywords, the conversion unit 12 will determine a priority for each conversion keyword converted from the input keyword to be used in generating hypothetical questions. The case where the input keyword can be converted into multiple conversion keywords means, for example, that the conversion model estimates multiple conversion keywords from one input keyword. For example, the conversion unit 12 will determine the priority so that conversion keywords that are easier for the question estimation model to ask questions about have a higher priority.

[0039] The transformation unit 12 determines the priority of transformation keywords, for example, based on the accuracy of the transformation model. The transformation unit 12 determines the priority of transformation keywords, for example, such that the priority of transformation keywords in the selected hypothetical question domain is higher than that of transformation keywords in the unselected hypothetical question domain. The transformation unit 12 determines the priority of transformation keywords, for example, such that the more frequently a transformation keyword is used, the higher its priority. The transformation unit 12 may also determine the priority of transformation keywords, for example, such that the more frequently a transformation keyword is used in a question sentence that uses the terminology corresponding to the training data, the higher its priority. The criteria for determining priority are set, for example, by the operator of the question estimation system 10.

[0040] The estimation unit 13 estimates the sentences of anticipated questions based on the converted keywords, using a question estimation model that estimates sentences of anticipated questions from keywords. The question estimation model is a learning model that estimates anticipated questions using the input terms. For example, the estimation unit 13 uses the converted keywords as input to the input question estimation model to estimate sentences of anticipated questions based on the converted keywords. Alternatively, the estimation unit 13 may use the input keywords as input to the question estimation model to estimate sentences of anticipated questions based on the converted keywords. Anticipated questions are, for example, questions that are likely to be asked in relation to the terms.

[0041] For example, suppose a user wants to obtain anticipated questions from the keyword "novel C virus." In this case, if "novel C virus" is a newly used term or a term whose frequency of use is increasing, the question estimation model is unlikely to be able to anticipate the questions. Therefore, the estimation unit 13 uses the conversion unit 12 to convert "novel C virus" into a keyword that the question estimation model can anticipate questions using. Then, the estimation unit 13 uses the converted "novel influenza" as input to the question estimation model to estimate anticipated questions corresponding to "novel influenza." In other words, when the question estimation model would normally estimate anticipated questions from "novel C virus," it uses the keyword "novel influenza," which has been converted from "novel C virus," to estimate anticipated questions. For example, if a question about sales may be asked in relation to the term "novel influenza," which has been converted from "novel C virus," the question estimation model estimates the anticipated question, "A new influenza is spreading, and consumption is falling. What measures are you considering?"

[0042] A question estimation model may estimate hypothetical questions containing multiple keywords from a set of multiple keywords. Alternatively, a question estimation model may estimate hypothetical questions containing some of the multiple keywords from a set of multiple keywords. Furthermore, a question estimation model may be a learning model that searches a database of hypothetical questions for hypothetical questions corresponding to keywords and uses the retrieved hypothetical questions as the estimation result.

[0043] The question estimation model is a learning model generated, for example, by learning the relationship between keywords and questions related to those keywords. The question estimation model is generated, for example, by well-known natural language processing techniques and deep learning using neural networks. The question estimation model is generated, for example, in a system outside of the question estimation system 10.

[0044] The question estimation model is generated using terms and sentences as training data that include at least one of the following: press releases, academic papers, newspaper articles, magazine articles, web articles, newsletters, Q&A collections, posts on social networking services (SNS), lecture proceedings, public announcements, and meeting minutes. For example, if the intended use of the anticipated questions is to prepare for questions that may be asked at a shareholders' meeting, the question estimation model is generated using terms and sentences as training data that include at least one of the following: Q&A sessions from past shareholders' meetings, anticipated Q&A collections, press releases, academic papers, newspaper articles, magazine articles, web articles, newsletters, financial results announcements, and IR (Investor Relations) information. The training data is not limited to the terms and sentences mentioned above.

[0045] The estimation unit 13 may estimate anticipated questions from the conversion keywords using one of a plurality of question estimation models. The user may select, for example, which of the plurality of conversion models to use. The estimation unit 13 may estimate multiple anticipated questions from one conversion keyword by using one conversion keyword as input to each of the plurality of conversion models. The estimation unit 13 may estimate anticipated questions from the conversion keywords using a question estimation model corresponding to the field of the anticipated questions selected by the user. For example, the estimation unit 13 estimates anticipated questions from the conversion keywords using a question estimation model corresponding to the field of the anticipated questions selected by the user.

[0046] The estimation unit 13 may estimate a hypothetical question using a selected conversion keyword from among several candidate conversion keywords. The selection of the conversion keyword is performed, for example, by the user. The estimation unit 13 may also estimate a hypothetical question using a keyword whose priority meets a predetermined criterion. For example, the estimation unit 13 estimates a hypothetical question using a keyword whose priority is above the criterion. The estimation unit 13 may also estimate a hypothetical question using a predetermined number of conversion keywords in descending order of priority. The priority is determined, for example, by the conversion unit 12. The predetermined criterion and predetermined number are set, for example, by the operator of the question estimation system 10.

[0047] The estimation unit 13 may estimate anticipated questions using keywords whose conversion accuracy in the conversion model meets a predetermined standard. For example, the estimation unit 13 estimates anticipated questions using keywords whose conversion accuracy in the conversion model is above the standard. Alternatively, the estimation unit 13 may estimate anticipated questions using a predetermined number of conversion keywords in descending order of conversion accuracy in the conversion model. The predetermined standard and predetermined number are set, for example, by the operator of the question estimation system 10.

[0048] The estimation unit 13 may modify the question estimation model to estimate the expected questions from the converted keywords if the question estimation model fails to estimate the expected questions from the converted keywords. The estimation unit 13 may also modify the question estimation model to estimate the expected questions from the converted keywords if the number of expected questions estimated by the question estimation model from the converted keywords is less than a certain threshold. The threshold for the number of expected questions that warrants modifying the question estimation model is set, for example, by the operator of the question estimation system 10. Furthermore, if the conversion unit 12 modifies the conversion model and re-converts the input keywords, the estimation unit 13 may, for example, use the re-converted converted keywords to estimate the expected questions.

[0049] If the question estimation model is unable to estimate a hypothetical question from the conversion keywords, the estimation unit 13 may input a different conversion keyword to the question estimation model and estimate the hypothetical question from the conversion keywords. The inability of the question estimation model to estimate a hypothetical question from the conversion keywords includes, for example, cases where the accuracy of the hypothetical question estimation by the question estimation model is below a certain standard. The accuracy of the hypothetical question estimation being below a certain standard means, for example, that the index indicating the accuracy of the estimation result is lower than the standard.

[0050] The estimation unit 13 may estimate hypothetical questions using the input keywords as input to a question estimation model, and if it is unable to estimate hypothetical questions, it may estimate hypothetical questions using the converted keywords as input to the question estimation model. Alternatively, the estimation unit 13 may estimate hypothetical questions using the input keywords as input to a question estimation model, and if the number of hypothetical questions estimated from the input keywords is less than a certain threshold, it may estimate hypothetical questions using the converted keywords as input to the question estimation model.

[0051] The estimation unit 13 may generate estimation results for questions with example answers added to them. When generating estimation results for questions with example answers added, the estimation unit 13 extracts example answers given for the same question from the database, for example. The estimation unit 13 also generates estimation results by associating the question with the example answers extracted from the database, for example. When example answers are added to a question, the database of example answers given for the question is generated in advance and stored in the storage unit 16.

[0052] The replacement unit 14 replaces the conversion keywords contained in the sentence of the hypothetical question estimated by the estimation unit 13 with the input keywords. For example, suppose the input keyword is "novel C virus", the conversion keyword is "novel influenza", and the hypothetical question estimated by the estimation unit 13 is "A novel influenza virus is spreading, and consumption is falling. What measures are you considering?" In this case, the replacement unit 14 replaces "novel influenza" with "novel C virus" in the hypothetical question estimated by the estimation unit 13. Then, the replacement unit 14 generates the sentence "A novel C virus is spreading, and consumption is falling. What measures are you considering?"

[0053] If the estimation result of the hypothetical question generated by the estimation unit 13 includes example answers, and the hypothetical question contains a conversion keyword, the replacement unit 14 replaces, for example, the conversion keyword included in the example answer with the input keyword.

[0054] The output unit 15 outputs the sentence of the assumed question that has been replaced by the replacement unit 14. The output unit 15 outputs the sentence of the assumed question that has been replaced by the replacement unit 14 to, for example, the terminal device 20. The output unit 15 may also output the estimation results to a display device (not shown) connected to the question estimation system 10. Alternatively, the output unit 15 may output the estimation results to a printing device connected via a network.

[0055] The output unit 15 may further output the conversion keywords used to estimate the sentence of the hypothetical question. For example, the output unit 15 may output the sentence of the hypothetical question in association with the conversion keywords used to estimate the sentence of the hypothetical question.

[0056] The output unit 15 may output the sentences of the assumed questions converted by the replacement unit 14 based on the priority of the conversion keywords. The output unit 15 outputs the sentences of the assumed questions converted by the replacement unit 14 in the form of a list arranged in order of the priority of the conversion keywords.

[0057] The output unit 15 may output the sentences of the assumed questions transformed by the replacement unit 14 based on the accuracy of the estimation of the question estimation model. The output unit 15 outputs the sentences of the assumed questions transformed by the replacement unit 14 in the form of a list arranged in order of the accuracy of the estimation of the question estimation model.

[0058] The output unit 15 may further output example answers to questions if example answers are attached to the estimation results of the hypothetical questions generated by the estimation unit 13. For example, the output unit 15 outputs the text of the hypothetical question converted by the replacement unit 14 in association with the example answers attached to the hypothetical question. If the example answer contains a conversion keyword, the output unit 15 outputs an example answer in which the conversion keyword has been replaced with the input keyword. Alternatively, the output unit 15 may output an example answer with the conversion keyword before it was replaced with the input keyword added.

[0059] If the user can select the conversion keyword, the output unit 15 may output candidate conversion keywords. The output unit 15 outputs candidate conversion keywords to, for example, the terminal device 20. The output unit 15 may also output each conversion keyword along with the field of the assumed question corresponding to that conversion keyword. For example, when outputting candidate conversion keywords converted from an input keyword containing the word "virus," the output unit 15 may output each candidate conversion keyword with information indicating whether it is a conversion keyword in the medical field or the information science field.

[0060] If the user can select a conversion model, the output unit 15 may output candidate conversion models. For example, the output unit 15 outputs candidate conversion models to the terminal device 20. Also, if the user can select a question estimation model, the output unit 15 may output candidate question estimation models. For example, the output unit 15 outputs candidate question estimation models to the terminal device 20.

[0061] Figure 4 shows an example of a display screen showing the estimated results of anticipated questions. In the example screen in Figure 4, the "Anticipated Questions" section displays the following questions: "With the spread of the novel coronavirus C, consumption is declining. What measures are you considering?" and "What is your business continuity plan in case the novel coronavirus C spreads?" Users of the question estimation system 10 can refer to the estimated results of anticipated questions, such as the example screen in Figure 4, to anticipate questions and prepare answers to those questions.

[0062] Figure 5 shows an example of a display screen that further displays conversion keywords in the estimated results of anticipated questions. In the example of the display screen in Figure 5, similar to the example of the display screen in Figure 4, the "Anticipated Question" display field shows the question content as "With the spread of the novel C virus, consumption is declining. What measures are you considering?" and "What is your business continuity plan in the event of a widespread outbreak of the novel C virus?". In addition, in the example of the display screen in Figure 5, the input keyword "novel C virus" is underlined, and the conversion word "novel influenza" is displayed below the underline. Users of the question estimation system 10 can, for example, refer to the estimated results of anticipated questions, such as the example of the display screen in Figure 5, to confirm the keywords used in estimating the anticipated questions. Users of the question estimation system 10 can, for example, refer to the estimated results of anticipated questions, such as the example of the display screen in Figure 5, to judge the accuracy of the estimated results of the anticipated questions.

[0063] Figure 6 shows an example of a display screen when example answers are displayed on the screen showing the estimated results of anticipated questions. In the example of the display screen in Figure 6, the estimated results of the anticipated questions are displayed in the "Anticipated Questions" column on the left, similar to the example of the display screen in Figure 4. In addition, in the example of the display screen in Figure 6, example answers corresponding to the anticipated questions are displayed in the "Example Answers" column on the right. In the example of the display screen in Figure 6, for the anticipated question, "With the spread of the novel coronavirus C, consumption is declining. What measures are you considering?", the example answer, "We plan to expand our product range for use in remote environments," is displayed. Also in the example of the display screen in Figure 6, for the anticipated question, "What is your business continuity plan in the event of a widespread outbreak of the novel coronavirus C?", the example answer, "We have created an environment where almost all employees can continue operations from home," is displayed. In the example of the display screen in Figure 6, one example answer is displayed for one question, but multiple example answers may be displayed for a single question. Users of the question estimation system 10 can more easily prepare answers to anticipated questions by referring to the estimated results of anticipated questions, such as the example display screen shown in Figure 6.

[0064] The memory unit 16 stores data used to generate hypothetical questions. The memory unit 16 stores input keywords acquired by the acquisition unit 11. The memory unit 16 stores, for example, the estimation results of hypothetical questions generated by the estimation unit 13. The memory unit 16 also stores, for example, the hypothetical questions after substitution by the substitution unit 14. The memory unit 16 stores, for example, a table showing the relationship between input keywords and conversion keywords. The memory unit 16 stores, for example, a conversion model. The memory unit 16 stores, for example, a question estimation model. The conversion model and the question estimation model may be stored in memory means other than the memory unit 16.

[0065] Terminal device 20 is, for example, a terminal device used by a user when using the question estimation system 10 to input input keywords and view the estimation results of anticipated questions. Terminal device 20 acquires input keywords used to generate anticipated questions, for example, which are entered by the user's operation. Then, terminal device 20 outputs the input keywords to the question estimation system 10, for example. Terminal device 20 acquires the estimation results of anticipated questions from the question estimation system 10, for example. Then, terminal device 20 displays the estimation results of anticipated questions on a display device (not shown), for example.

[0066] When a conversion model is selected, the terminal device 20 acquires the selection of the conversion model input by the operator. The terminal device 20 then outputs the acquired selection of the conversion model to the question estimation system 10. Similarly, when a question estimation model is selected, the terminal device 20 acquires the selection of the question estimation model input by the operator. The terminal device 20 then outputs the acquired selection of the question estimation model to the question estimation system 10.

[0067] Multiple terminal devices 20 may be connected to the question estimation system 10. When a question estimation model is selected, at least one of the conversion model and the question estimation model used to generate anticipated questions may be set for each user logged into the question estimation system 10, or for each terminal device 20.

[0068] For example, a personal computer may be used as the terminal device 20. The terminal device 20 may also be a smartphone or a tablet computer. The terminal device 20 is not limited to the above.

[0069] The operation of the question estimation system 10 will now be explained. Figure 7 shows an example of the operation flow when the question estimation system 10 estimates a hypothetical question.

[0070] The acquisition unit 11 acquires the keywords of the anticipated questions to be generated as input keywords (step S11). The acquisition unit 11 acquires the input keywords, for example, from the terminal device 20.

[0071] When an input keyword is obtained, the conversion unit 12 converts the input keyword into a conversion keyword which is a synonym or similar keyword (step S12). The conversion unit 12 converts the input keyword into a conversion keyword using, for example, a conversion model. The conversion model is, for example, a learning model that converts input terms into synonymous or similar terms. The conversion unit 12 converts the input keyword into a conversion keyword using, for example, the input keyword as input to the conversion model.

[0072] When the input keywords are converted into conversion keywords, the estimation unit 13 estimates the sentence of the anticipated question using a question estimation model that estimates the sentence of the anticipated question from the input terms, based on the conversion keywords (step S13). The estimation unit 13 uses the conversion keywords as input to the question estimation model and estimates the anticipated question from the conversion keywords.

[0073] If the estimated hypothetical question satisfies the criteria (Yes in step S14), the replacement unit 14 replaces the conversion keywords contained in the sentence of the hypothetical question estimated by the estimation unit 13 with the input keywords (step S15). The criteria is, for example, a criterion for determining whether to convert the input keyword to a keyword. Also, if the input keyword is a conversion keyword, the conversion criterion is, for example, a criterion for determining whether to re-convert the input keyword to the conversion keyword. In other words, for the hypothetical question estimated by the question estimation model to satisfy the criteria is, for example, that the question estimation model can estimate the hypothetical question. Also, for the hypothetical question estimated by the question estimation model to satisfy the criteria is that the number of hypothetical questions estimated by the question estimation model may be greater than or equal to a predetermined number. The above is not the only way in which the estimated hypothetical question satisfies the criteria.

[0074] When the text of the anticipated question is replaced with the conversion keywords contained in it, the output unit 15 outputs the text of the anticipated question that has been replaced by the replacement unit 14 (step S16). The output unit 15 outputs the text of the anticipated question that has been replaced by the replacement unit 14 to, for example, the terminal device 20.

[0075] Furthermore, in step S14, if the estimated hypothetical question does not meet the criteria (No in step S14), the replacement unit 14 returns to step S12, and the transformation unit 12 performs the process of converting the input keyword to a transformation keyword. The replacement unit 14 converts the input keyword to a transformation keyword that is different from the case where the estimated hypothetical question did not meet the criteria. The replacement unit 14 converts the input keyword to a transformation keyword by changing the transformation model, for example.

[0076] The question estimation system 10 obtains, for example, keywords that the user wants to use to create a hypothetical question as input keywords. The question estimation system 10 converts the obtained input keywords into conversion keywords suitable for input to the question estimation model. The question estimation system 10 uses the conversion keywords as input to estimate a hypothetical question using the question estimation model. Then, the question estimation system 10 replaces the conversion keywords contained in the estimated hypothetical question sentence with the input keywords. In this way, by estimating a hypothetical question after converting the input keywords in the question estimation system 10, the user of the question estimation system 10 does not need to be aware of, for example, whether the input keywords are suitable as input to the question estimation model. For example, even if the user of the question estimation system 10 freely selects keywords for which they want to create a hypothetical question, there is a high probability that they will obtain a hypothetical question corresponding to those keywords. For this reason, by using the question estimation system 10, it is possible to easily estimate questions corresponding to keywords.

[0077] Furthermore, by using a conversion model to convert the input keywords, the question estimation system 10 can convert the input keywords into conversion keywords that are more suitable for the input of the question estimation model. For example, even if a newly coined term is input as an input keyword, the conversion model can be used to convert it into a conversion keyword that is semantically similar to the input keyword and is more suitable for the input of the question estimation model. Therefore, even if a newly coined term is input as a keyword, the question estimation system 10 can use the question estimation model to estimate an appropriate anticipated question.

[0078] Each process in the question estimation system 10 may be distributed and executed across multiple information processing devices connected via a network. For example, the processes in the conversion unit 12 and the substitution unit 14, and the processes in the estimation unit 13 may be performed on separate information processing devices. The choice of which of the multiple information processing devices performs each process in the question estimation system 10 can be set as appropriate.

[0079] Each process in the question estimation system 10 can be implemented by executing a computer program on a computer. Figure 8 shows an example of the configuration of a computer 100 that executes the computer programs that perform each process in the question estimation system 10. The computer 100 includes a CPU (Central Processing Unit) 101, memory 102, storage device 103, input / output interface (I / F) 104, and communication interface (I / F) 105.

[0080] The CPU 101 reads and executes computer programs that perform various processes from the storage device 103. The CPU 101 may be composed of a combination of multiple CPUs. Alternatively, the CPU 101 may be composed of a combination of a CPU and another type of processor. For example, the CPU 101 may be composed of a combination of a CPU and a GPU (Graphics Processing Unit). The memory 102 is composed of DRAM (Dynamic Random Access Memory) or the like, and temporarily stores computer programs executed by the CPU 101 and data being processed. The storage device 103 stores computer programs executed by the CPU 101. The storage device 103 is composed of, for example, a non-volatile semiconductor storage device. Other storage devices such as hard disk drives may be used for the storage device 103. The input / output I / F 104 is an interface that receives input from the operator and outputs display data, etc. The communication I / F 105 is an interface that sends and receives data with terminal devices or other information processing devices. The terminal device 20 can also be configured similarly.

[0081] The computer programs used to execute each process can also be stored and distributed on a computer-readable recording medium that non-temporarily stores data. Examples of recording media include magnetic tapes for data recording and magnetic disks such as hard disks. Optical discs such as CD-ROMs (Compact Disc Read Only Memory) can also be used as recording media. Non-volatile semiconductor memory devices may also be used as recording media.

[0082] Some or all of the above embodiments may also be described as follows, but are not limited to the following:

[0083] [Note 1] A means for obtaining keywords for the anticipated questions to be generated as input keywords, A conversion means that converts the aforementioned input keyword into a conversion keyword which is a synonym or similar keyword, An estimation means for estimating the text of a hypothetical question using a question estimation model that estimates the text of a hypothetical question from the keywords based on the aforementioned conversion keywords, A replacement means that replaces the conversion keyword contained in the sentence of the hypothetical question estimated by the estimation means with the input keyword, The output means outputs the sentence of the hypothetical question that has been replaced by the substitution means. A question estimation system equipped with the following features.

[0084] [Note 2] The conversion means converts the input keyword to the converted keyword using a conversion model that converts the input term to a synonymous or similar term. The question estimation system described in Appendix 1.

[0085] [Note 3] The conversion means converts the input keyword into a converted keyword that enables the question estimation model to estimate the sentence of the anticipated question. The question estimation system described in Appendix 1 or 2.

[0086] [Note 4] The conversion means converts the input keyword to the conversion keyword when the input keyword satisfies a criterion set based on the novelty of the term. A question estimation system described in any of the appendices 1 to 3.

[0087] [Note 5] The conversion means converts the input keyword to the conversion keyword if the estimation means is unable to estimate the expected question from the input keyword. A question estimation system described in any of the appendices 1 to 3.

[0088] [Note 6] The conversion means determines the priority to be used for estimating the anticipated question for each conversion keyword, The estimation means estimates the text of the anticipated question using the question estimation model based on the conversion keyword whose priority satisfies the criteria. A question estimation system described in any of the appendices 1 through 5.

[0089] [Note 7] The conversion means converts the input keywords into conversion keywords using a conversion model corresponding to the field of the assumed question. A question estimation system described in any of the appendices 1 through 6.

[0090] [Note 8] The output means further outputs the conversion keywords used to estimate the sentence of the assumed question. A question estimation system described in any of the appendices 1 through 7.

[0091] [Note 9] The output means outputs candidate conversion keywords corresponding to the input keyword, The estimation means estimates the text of the hypothetical question using the question estimation model based on the conversion keyword selected from the candidate conversion keywords. A question estimation system described in any of the appendices 1 through 8.

[0092] [Note 10] The output means outputs the sentence of the assumed question that has been converted by the conversion means, based on the priority of the conversion keywords. A question estimation system described in any of the appendices 1 through 9.

[0093] [Note 11] The aforementioned anticipated questions are those that are expected to be asked by attendees at the shareholders' meeting. A question estimation system described in any of the appendices 1 through 10.

[0094] [Note 12] The keywords for the anticipated questions to be generated are obtained as input keywords, The aforementioned input keyword is converted into a conversion keyword which is a synonym or similar keyword. Based on the aforementioned conversion keywords, a question estimation model is used to estimate the text of the anticipated question from the keywords. Replace the conversion keywords contained in the estimated hypothetical question text with the input keywords, Output the substituted text of the hypothetical question. Question estimation method.

[0095] [Note 13] The process involves obtaining keywords for the anticipated questions to be generated as input keywords, A process to convert the aforementioned input keyword into a conversion keyword which is a synonym or similar keyword, Based on the aforementioned conversion keywords, a question estimation model is used to estimate the text of the anticipated question from the keywords, and the process involves estimating the text of the anticipated question. A process of replacing the conversion keywords contained in the estimated hypothetical question text with the input keywords, The process of outputting the substituted sentences of the hypothetical questions and A recording medium that non-temporarily stores a question estimation program that causes a computer to run.

[0096] The present disclosure has been explained above using the embodiments described in the preceding text as examples. However, the present disclosure is not limited to the embodiments described above. That is, the present disclosure can be applied in various forms that can be understood by those skilled in the art within the scope of the present disclosure. [Explanation of Symbols]

[0097] 10-Question Estimation System 11 Acquisition Department 12 Conversion section 13 Estimation part 14 Replacement part 15 Output section 16 Memory section 20 Terminal devices 100 Computers 101 CPU 102 memory 103 Storage device 104 Input / Output Interfaces 105 Communication I / F

Claims

1. A means for obtaining keywords for the anticipated questions to be generated as input keywords, A conversion means that converts the aforementioned input keyword into a conversion keyword which is a synonym or similar keyword, An estimation means for estimating the text of a hypothetical question using a question estimation model that estimates the text of a hypothetical question from the keywords based on the aforementioned conversion keywords, A replacement means that replaces the conversion keyword contained in the sentence of the hypothetical question estimated by the estimation means with the input keyword, The output means outputs the sentence of the hypothetical question that has been replaced by the substitution means. A question estimation system equipped with the following features.

2. The conversion means converts the input keyword to the converted keyword using a conversion model that converts the input term to a synonymous or similar term. The question estimation system according to claim 1.

3. The conversion means converts the input keyword into a converted keyword that enables the question estimation model to estimate the sentence of the anticipated question. The question estimation system according to claim 1.

4. The conversion means converts the input keyword to the conversion keyword when the input keyword satisfies a criterion set based on the novelty of the term. The question estimation system according to claim 1.

5. The conversion means converts the input keyword to the conversion keyword if the estimation means is unable to estimate the expected question from the input keyword. The question estimation system according to claim 1.

6. The conversion means determines the priority to be used for estimating the anticipated question for each conversion keyword, The estimation means estimates the text of the anticipated question using the question estimation model based on the conversion keyword whose priority satisfies the criteria. The question estimation system according to claim 1.

7. The conversion means converts the input keywords into conversion keywords using a conversion model corresponding to the field of the assumed question. The question estimation system according to claim 1.

8. The output means further outputs the conversion keywords used to estimate the sentence of the assumed question. A question estimation system according to any one of claims 1 to 7.

9. A computer, The keywords for the anticipated questions to be generated are obtained as input keywords, The aforementioned input keyword is converted into a conversion keyword which is a synonym or similar keyword. Based on the aforementioned conversion keywords, a question estimation model is used to estimate the text of the anticipated question from the keywords. Replace the conversion keywords contained in the estimated hypothetical question text with the input keywords, Output the substituted text of the hypothetical question. Question estimation method.

10. The process involves obtaining keywords for the anticipated questions to be generated as input keywords, A process to convert the aforementioned input keyword into a conversion keyword which is a synonym or similar keyword, Based on the aforementioned conversion keywords, a question estimation model is used to estimate the text of the anticipated question from the keywords, and the process involves estimating the text of the anticipated question. A process of replacing the conversion keywords contained in the estimated hypothetical question text with the input keywords, The process of outputting the substituted sentences of the hypothetical questions and A question estimation program that has a computer perform the following actions.

Images