Information processing method and information processing system

Abstract

To estimate the uniqueness of an answer with higher accuracy.SOLUTION: An information processing system acquires a first vector corresponding to a first text showing a first answer to a subject, derives a first subscore corresponding to the height of uniqueness on the basis of similarity between a subject vector corresponding to a subject text showing the subject and the first vector or first similarity between a reference vector corresponding to a reference text showing an answer associated with the subject and the first vector, acquires, for the subject, a second vector corresponding to a past text showing a past answer made before the first answer, derives a second subscore on the basis of second similarity showing similarity between the first vector and the second vector, and derives a score showing the height of uniqueness of the first answer on the basis of the lower subscore between the first subscore and the second subscore.SELECTED DRAWING: Figure 5

Description

[Technical Field]

[0001] The present disclosure relates to techniques for evaluating people's ideas. [Background technology]

[0002] Non-Patent Document 1 discloses a technique for calculating a score of the uniqueness of a current answer to a topic. Specifically, Non-Patent Document 1 calculates a first sub-score indicating the low similarity between the topic and the current answer and a second sub-score indicating the low similarity between a past answer and the current answer, and linearly combines the first sub-score and the second sub-score to calculate the score of the current answer. [Prior art documents] [Non-patent literature]

[0003] [Non-Patent Document 1] David Fisher, Rishabh Choudhary, Omar Alsayed, Simona Doboli, Ali A.Minai, “Building Semantic Cognitive Maps with Text Embedding and Clustering”, 2022 International Joint Conference on Neural Networks (IJCNN), pp.01-08, 2022. Summary of the Invention [Problem to be solved by the invention]

[0004] However, in Non-Patent Document 1, the score of an answer is derived by linearly combining the first sub-score and the second sub-score, and therefore there is a problem in that the uniqueness of an answer cannot be estimated with high accuracy.

[0005] The present disclosure has been made in consideration of such problems, and aims to provide a technology for more accurately estimating the uniqueness of answers. [Means for solving the problem]

[0006] An information processing method in one embodiment of the present disclosure is an information processing method in a computer, which includes: obtaining a first vector corresponding to first text indicating a first answer to a topic; deriving a first sub-score corresponding to a level of uniqueness based on a first similarity indicating the similarity between the topic vector corresponding to the topic text indicating the topic and the first vector, or the similarity between the first vector and a reference vector corresponding to reference text indicating an answer associated with the topic; obtaining a second vector corresponding to past text indicating an answer to the topic given earlier than the first answer; deriving a second sub-score based on a second similarity indicating the similarity between the first vector and the second vector; and deriving a score indicating the uniqueness of the first answer based on the lower sub-score of the first sub-score and the second sub-score. [Effects of the Invention]

[0007] According to the present disclosure, the uniqueness of an answer can be estimated with greater accuracy. [Brief explanation of the drawings]

[0008] [Figure 1] FIG. 1 is a diagram conceptually illustrating the contents of the present embodiment. [Figure 2] FIG. 1 is a diagram of a booth to which an information processing system according to an embodiment is applied. [Figure 3] A diagram of a booth during a conference. [Figure 4] 10A and 10B are diagrams showing details of a theme input screen and a score screen. [Figure 5] 1 is a block diagram illustrating an example of a configuration of an information processing system according to an embodiment. [Figure 6] FIG. 2 is a diagram showing the data structure of a subject database and an answer database. [Figure 7] FIG. 2 is a diagram showing the data configuration of a conference database and a user database. [Figure 8]FIG. 10 is a diagram showing a lighting device installed in a booth. [Figure 9] FIG. 10 is a diagram illustrating a calculation process of a first vector. [Figure 10] 10 is a flowchart illustrating an example of a process when the information processing system acquires an answer. [Figure 11] 10 is a flowchart illustrating an example of processing when the information processing system evaluates an answer. [Figure 12] FIG. 10 is an explanatory diagram of a process for deriving a score. DETAILED DESCRIPTION OF THE INVENTION

[0009] (Findings underlying this disclosure) In research into assessing people's emotions and thoughts, methods are being considered to evaluate the originality of answers (ideas) to a certain topic. The similarity between a topic and an answer is calculated from the vector corresponding to the topic and the vector corresponding to the answer, and the lower the similarity, the higher the originality can be evaluated.

[0010] However, with this method, even if a response is similar to a previous response given by another person in a meeting with multiple participants, the response will be evaluated as highly original if the similarity to the topic is low.An answer that is similar to a response given by another person is a copy of another person's answer, and therefore cannot be considered highly original.

[0011] The conventional technology disclosed in Non-Patent Document 1 derives a final score by linearly combining the first sub-score and the second sub-score. In this way, even if an answer is similar to a past answer, if the answer is not similar to the topic, a higher score is derived for that answer. Also, in the conventional technology, even if an answer is similar to the topic, if the answer is not similar to the past answer, a higher score is derived for that answer. As such, the conventional technology averages the first sub-score and the second sub-score to derive the final score, which is insufficient for accurately determining originality.

[0012] Therefore, the inventors discovered that the uniqueness of an answer can be evaluated more accurately by deriving the lower of the first and second subscores as the score of the answer, and came up with the present disclosure.

[0013] (1) An information processing method in one aspect of the present disclosure is an information processing method in a computer, including: obtaining a first vector corresponding to a first text indicating a first answer to a topic; deriving a first sub-score corresponding to a level of uniqueness based on a first similarity indicating a similarity between the first vector and a topic vector corresponding to topic text indicating the topic, or a first similarity indicating a similarity between the first vector and a reference vector corresponding to reference text indicating an answer associated with the topic; obtaining a second vector corresponding to a past text indicating an answer to the topic given earlier than the first answer; deriving a second sub-score based on a second similarity indicating a similarity between the first vector and the second vector; and deriving a score indicating a level of uniqueness of the first answer based on the lower sub-score of the first sub-score and the second sub-score.

[0014] According to this configuration, the score of an answer is derived based on the lower of the first and second subscores. Therefore, even if either the first or second subscore is high, the score of the answer will not be influenced by the higher subscore. As a result, answers that are not similar to either past answers or the topic (or answers associated with the topic) can be given a high score, and the originality of the answer can be evaluated more accurately. Furthermore, by using the similarity between the first vector and a reference vector indicating an answer associated with the topic as the first similarity, a high score can be given to an answer that is more original.

[0015] (2) In the information processing method described in (1) above, the first similarity and the second similarity may each correspond to a cosine similarity.

[0016] According to this configuration, the cosine similarity can be used to derive the similarity with high accuracy.

[0017] (3) In the information processing method described in (1) or (2) above, when the score satisfies a first condition indicating low uniqueness, the method may further include notifying the user who provided the first answer of the reason for the low score.

[0018] According to this configuration, the user is notified of the reason for the low score, which encourages the user to change their state of mind and improves the efficiency of idea generation.

[0019] (4) In the information processing method described in (3) above, the notification of the cause may include notifying that the first answer is similar to a past answer of a specific person.

[0020] This configuration allows users to confirm that the low score of their answer is due to its similarity to the answer of a specific person. Furthermore, it is possible to understand the user's characteristics, such as whether they agree with a specific person.

[0021] (5) In the information processing method described in (3) above, the notification of the cause may include notifying that the first answer is not unique.

[0022] This configuration makes it possible to directly inform the user that the answer is not unique.

[0023] (6) In the information processing method described in (3) above, the notification of the cause may include notifying that the first answer is close to any of a past answer, the subject, and an answer associated with the subject.

[0024] This configuration makes it possible to specifically inform the user of the reason for the low score.

[0025] (7) In the information processing method described in any one of (1) to (6) above, when the score satisfies a first condition indicating low uniqueness, a lighting device that illuminates the vicinity of the responding user may be controlled to lower the color temperature of the lighting light and increase the illuminance.

[0026] It is known that lowering the color temperature of the illumination light and increasing the illuminance enhances the creativity of the user. With this configuration, by using illumination light for a user who is thought to be unable to demonstrate his or her creativity, the creativity of the user can be stimulated.

[0027] (8) In the information processing method described in any one of (1) to (7) above, the past text may be linked to the subject and stored in a storage device.

[0028] According to this configuration, it is possible to directly recognize which topic a past answer is an answer to, thereby reducing the processing load on the computer.

[0029] (9) In the information processing method described in any one of (1) to (8) above, the first vector may be a word vector among a plurality of word vectors corresponding to each of a plurality of words included in the first text, which has the smallest similarity to the thematic vector or the reference vector, and the second vector may be a word vector among a plurality of word vectors corresponding to each of a plurality of words included in the past text, which has the smallest similarity to the thematic vector or the reference vector.

[0030] According to this configuration, when the first text and the past text contain multiple words, the features of the first text and the past text can be appropriately expressed using word vectors, which allows the computer to accurately derive the first and second similarities and improve the processing capabilities of the computer.

[0031] (10) In another aspect of the present disclosure, an information processing system is an information processing system including a computer, wherein the computer includes a processor that executes the following processes: acquires a first vector corresponding to first text indicating a first answer to a topic; derives a first sub-score corresponding to a level of uniqueness based on a similarity between a topic vector corresponding to topic text indicating the topic and the first vector, or a first similarity indicating a similarity between a reference vector corresponding to reference text indicating an answer associated with the topic and the first vector; acquires a second vector corresponding to past text indicating an answer to the topic given earlier than the first answer; derives a second sub-score based on a second similarity indicating a similarity between the first vector and the second vector; and derives a score indicating a level of uniqueness of the first answer based on the lower sub-score of the first sub-score and the second sub-score.

[0032] This configuration provides an information processing system that can more accurately evaluate the uniqueness of answers.

[0033] The present disclosure can also be realized as an information processing program that causes a computer to execute each of the characteristic configurations included in such an information processing method. Needless to say, such a computer program can be distributed on a computer-readable non-transitory recording medium such as a CD-ROM or via a communication network such as the Internet.

[0034] Note that each of the embodiments described below represents a specific example of the present disclosure. The numerical values, shapes, components, steps, and step orders shown in the following embodiments are merely examples and are not intended to limit the present disclosure. Furthermore, among the components in the following embodiments, components that are not described in the independent claims that represent the highest concept are described as optional components. Furthermore, in all of the embodiments, the respective contents can be combined.

[0035] (Embodiment) FIG. 1 is a diagram conceptually illustrating the contents of this embodiment. In FIG. 1, each circle indicates the position of a vector in vector space. The topic vector 31 is a vector corresponding to the topic text indicating the topic. The first vector 10 is a vector corresponding to the first text indicating the latest answer to the topic. The second vectors 21, 22, and 23 are vectors corresponding to past answers to the topic, respectively.

[0036] A theme is a question that requires a user to come up with a highly original idea. For example, a theme is a question such as, "Could you give me a unique way to use a certain object (e.g., a scrubbing brush)?" An answer is an idea that a user comes up with in response to the theme. In this embodiment, the originality of this answer is quantitatively evaluated.

[0037] Similarity a is the similarity between the first vector 10 and the thematic vector 31. Similarity b1 is the similarity between the first vector 10 and the second vector 21, similarity b2 is the similarity between the first vector 10 and the second vector 22, and similarity b3 is the similarity between the first vector 10 and the second vector 23.

[0038] The first subscore A is a score that evaluates the originality of the answer based on the similarity a. The second subscores B1, B2, and B3 are scores that evaluate the originality of the answer based on the similarities b1, b2, and b3, respectively. The first subscore A and the second subscores B1, B2, and B3 have higher scores as the similarities a, b1, b2, and b3, respectively, are lower.

[0039] The conventional technology shown in Non-Patent Document 1 derives a final score for evaluating the uniqueness of an answer by linearly combining (weighted addition) the first sub-score A and the second sub-scores B1, B2, and B3. In contrast, this embodiment compares the first sub-score A with the second sub-score B3, which has the least uniqueness among the second sub-scores B1, B2, and B3, and derives a final score based on the second sub-score B3, which has the smallest score.

[0040] FIG. 2 is a diagram of a booth 400 to which the information processing system 1 according to the embodiment is applied. The booth 400 is a space such as a room where one or more users work. The booth 400 is, for example, a conference room. The work may be, for example, a meeting to generate ideas, such as brainstorming. The booth 400 is equipped with a display device 102 as an output unit and an input device 101 as an input unit. In this example, two users 401 and 402 reside in the booth 400. The display device 102 is hung on, for example, a wall of the booth 400 so that the users 401 and 402 can easily share their screens. The input device 101 is installed on a desk installed in the booth 400. The input device 101 includes, for example, a microphone 101a, a keyboard 101b, and a mouse 101c. The microphone 101a picks up responses spoken by the users 401 and 402 to a topic. The keyboard 101b accepts text input indicating the responses of the users 401 and 402 to the topic. The mouse 101c accepts operations by users 401 and 402.

[0041] The display device 102 displays a topic input screen 510 if the meeting has not yet started, and displays a score screen 520 (see FIG. 3) including the topic and the score for the previous answer if the meeting is in progress.

[0042] Users 401 and 402 can set a topic and start a conference using the input device 101. In Fig. 2, there are two users 401 and 402 in the booth 400, but the number of users may be three or more, or may be one.

[0043] The screen displayed by the display device 102 may be displayed on the display of a mobile terminal (such as a notebook computer) owned by the users 401 and 402.

[0044] The conference may be an online conference. In this case, booth 400 is made up of multiple booths used by multiple users participating in the online conference. Booth 400 may also be used for a hybrid conference in which users 401 and 402 and another user participate online. Display device 102 may be installed on a desk, or may be made up of a display unit of a mobile terminal owned by users 401 and 402.

[0045] 3 is a diagram of the booth 400 during a conference. After the conference starts, the display device 102 displays a score screen 520 including a topic. The users 401 and 402 use the input device 101 to provide answers to the topic. The score screen 520 then displays the scores for the answers.

[0046] The input device 101 may set a break between answers based on, for example, a silent period. For example, the input device 101 may determine that one answer to a topic has ended when a silent period of five seconds has continued. The silent period of five seconds is an example, and any appropriate value of four seconds or less, six seconds or more may be used.

[0047] 4 is a diagram showing details of the topic input screen 510 and the score screen 520. The topic input screen 510 includes a topic input field 511 and a start meeting button 512. The topic input field 511 displays a topic that the user inputs using the input device 101. The user can input the topic by voice or using the keyboard 102b.

[0048] The conference start button 512 is a button that the user presses to notify the information processing system 1 of the start of a conference. When the user inputs a subject in the subject input field 511 and presses the conference start button 512, the conference starts.

[0049] The score screen 520 includes a subject display field 521, a score display field 522, and an end meeting button 523. The subject display field 521 displays the subject entered by the user in the subject input field 511. In this example, the subject display field 521 displays "A new way to use ping pong balls."

[0050] The score display field 522 displays the score of the most recent response to the topic. In this example, the score display field 522 displays "Current utterance score: 80 points." The end conference button 523 is a button that is pressed by the user when ending the conference. When this button is pressed, the information processing system 1 recognizes that the conference has ended. When the information processing system 1 accepts an operation to select the end conference button, it displays the topic input screen 510 on the display device 102. This returns the display on the display device 102 to its initial state.

[0051] The topic input screen 510 may include an answer start button and an answer end button. The answer start button is a button for the user to notify the information processing system 1 that he or she is about to answer the topic. The answer end button is a button for the user to notify the information processing system 1 that he or she has finished answering the current topic. In this case, the user uses the input device 101 to input the operation of pressing the answer start button and the answer end button each time he or she answers. In a conference, various conversations other than answers may take place. Such conversations include, for example, meaningless conversations such as chatter, conversations to explain the topic, and conversations to determine the topic. By providing the answer start button and the answer end button, the information processing system 1 does not need to process conversations other than answers, thereby reducing the processing load on the information processing system 1.

[0052] 5 is a block diagram showing an example of the configuration of an information processing system 1 according to an embodiment. The information processing system 1 includes an input device 101, a display device 102, a database device 200, a speech processing device 300, and a lighting device 500. The input device 101, the display device 102, and the lighting device 500 are installed in a booth 400. The input device 101 includes a communication unit. The display device 102 includes a communication unit. Details of the input device 101 and the display device 102 have been described above, so further explanation will be omitted.

[0053] The database device 200 and the speech processing device 300 are configured, for example, by a cloud server. However, this is just an example, and the database device 200 and the speech processing device 300 may be configured by an edge computer. In the example of FIG. 5, the database device 200 and the speech processing device 300 are illustrated as separate configurations, but this is just an example, and they may be configured by the same cloud server or edge computer. The input device 101 and the display device 102 are communicably connected to the database device 200 via a communication network NT1. The display device 102 and the speech processing device 300 are communicably connected to each other via a communication network NT2. The database device 200 and the speech processing device 300 are communicably connected to each other via a communication network NT3. The speech processing device 300 and the lighting device 500 are communicably connected to each other via a communication network NT4.

[0054] The communication networks NT1, NT2, NT3, and NT4 are configured by wide area communication networks including, for example, the Internet and a mobile phone communication network. When the database device 200 and the database device 200 are configured by edge computers, the communication networks NT1 to NT4 are configured by local area networks.

[0055] The speech processing device 300 includes a processor 310, a communication unit 320, and a memory 330. The processor 310 is configured, for example, by a central processing unit (CPU). The processor 310 includes a display control unit 301, a score calculation unit 302, and a control unit 303. The display control unit 301 to the control unit 303 are realized by the CPU executing an information processing program. However, this is just an example, and the display control unit 301 to the control unit 303 may also be configured by dedicated integrated circuits.

[0056] The score calculation unit 302 acquires a first vector corresponding to a first text indicating a first answer to a topic. Specifically, the score calculation unit 302 acquires the first text from the answer database 202 using the communication unit 320. The score calculation unit 302 acquires the first vector by inputting the acquired first text into a language model.

[0057] The language model can be a model that has been machine-learned in advance to convert text into word vectors. Word2vec can be used as the word vector. Therefore, the first vector is represented using word2vec. BERT (Bidirectional Encoder Representations from Transformers) can also be used as the language model.

[0058] The score calculation unit 302 acquires the topic text from the topic database 201 using the communication unit 320, and converts the acquired topic text into word vectors in the same way as the first text, thereby acquiring the topic vectors.

[0059] The score calculation unit 302 derives a first sub-score corresponding to the level of uniqueness based on a first similarity indicating the similarity between the thematic vector and the first vector. The similarity may be, for example, a cosine similarity. However, this is just one example, and the similarity may also be a Pearson correlation coefficient. The same applies to the second similarity described below.

[0060] The lower the first similarity, the larger the value of the first subscore. The first subscore can be expressed, for example, as 1-cosine similarity. Cosine similarity takes a value between -1 and 1, with values ​​closer to -1 indicating lower similarity and closer to 1 indicating higher similarity. By using 1-cosine similarity as the first subscore, the first subscore takes a value between 0 and 2. Therefore, the lower the similarity (the higher the uniqueness), the closer the first subscore is to 2, and the higher the similarity (the lower the uniqueness), the closer it is to 0. The same applies to the second subscore, which will be described later.

[0061] The score calculation unit 302 may derive the first similarity as the similarity between the first vector and a reference vector corresponding to reference text indicating an answer associated with the topic. An answer associated with the topic is an answer associated with the topic that is considered to have low uniqueness. An answer that is recognized as commonplace by many people, such as the answer "lay it on the floor" to the topic "how to use a futon," corresponds to an answer associated with the topic. As a result, if the first answer is not similar to the topic but is similar to an answer recognized as commonplace, the first similarity between the first vector and the reference vector becomes high, thereby preventing a high score from being derived for such a first answer. In this case, the score calculation unit 302 may record reference text or reference vectors corresponding to the topic in memory 330 in advance and derive the first similarity using the reference text or reference vector.

[0062] The score calculation unit 302 acquires a second vector corresponding to past text indicating an answer given to the topic earlier than the first answer. Specifically, the score calculation unit 302 acquires the past text from the answer database 202 using the communication unit 320, converts the acquired past text into a word vector in the same way as the first text, and acquires the second vector.

[0063] The score calculation unit 302 derives a second sub-score based on a second similarity indicating the similarity between the first vector and the second vector.

[0064] The score calculation unit 302 derives a score indicating the degree of uniqueness of the first answer based on the lower of the first subscore and the second subscore. In particular, the score calculation unit 302 derives the lower of the first subscore and the second subscore as the score. A value normalized within a predetermined range (e.g., 0 to 100) may be used as the score. Note that the score calculation unit 302 may derive a score indicating the degree of uniqueness of the first answer based on the higher of the first similarity and the second similarity.

[0065] The display control unit 301 generates display data for the subject input screen 510 and the score screen 520, and transmits the display data to the display device 102 using the communication unit 320, thereby displaying the subject input screen 510 and the score screen 520 on the display device 102. The display control unit 301 only needs to generate display data for displaying the score derived by the score calculation unit 302 on the score screen 520.

[0066] When the score indicates a first condition indicating low uniqueness, the display control unit 301 may notify the user who provided the first answer of the reason for the low score. The first condition may be, for example, that the score is lower than a threshold. Alternatively, the first condition may be that the number of times the score is lower than the threshold is a predetermined number of times in a row.

[0067] The notification of the cause may include notifying the user that the first answer is similar to a past answer of a specific person. The display control unit 301 identifies, from the answer database 202 described below, past answers of one or more second users different from the first user that have the highest similarity to the first answer of the first user among past answers given during the same conference. The display control unit 301 tallies the number of identified past answers for each of one or more second users. The display control unit 301 identifies a second user whose number of tallies is equal to or greater than a reference number as a third user. The display control unit 301 may also identify a second user whose number of tallies is equal to or greater than the reference number and has the largest number of tallies as the third user. The display control unit 301 displays a message on the score screen 520 indicating that the first answer is similar to the answer of the third user.

[0068] Referring to FIG. 6, assume that the answer with answer ID "8" is the first answer. The answer with answer ID "8" was spoken by a user with user ID "2," so this user is the first user. The display control unit 301 extracts records that record answer texts of the user with user ID "2" from the answer database 202. Here, five records are extracted. The display control unit 301 references the similar answer fields of the five extracted records and identifies answer IDs of past answers of users other than user ID "2" that are most similar to the answer of user ID "2." Here, three answer IDs "1," one answer ID "5," and one answer ID "6" are identified. The display control unit 301 tallies the number of past answers of users other than user ID "2" from the identified answer IDs, by user. In this example, answer ID "1," answer ID "5," and answer ID "6" were all spoken by the user with user ID "1," so the number of past answers for user ID "1" is tallied as "5." If the reference number is 5, then the number of past answers for user ID "1," "5," is greater than or equal to the reference number, so user ID "1" is identified as the third user.

[0069] The notification of the cause may include notifying that the first answer is not unique. The display control unit 301 may display a message indicating that the first answer is not unique on the score screen 520.

[0070] The notification of the cause may include notification that the first answer is close to any of a past answer, the main topic, and an answer associated with the main topic. When the score calculation unit 302 selects the second sub-score low to derive a score, the display control unit 301 determines that the first answer is close to the past answer. In this case, the display control unit 301 may display a message indicating that the first answer is close to the past answer on the score screen 520. When the score calculation unit 302 selects the first sub-score low to derive a score, the display control unit 301 may determine that the first answer is close to the main topic. In this case, the display control unit 301 may display a message indicating that the first answer is close to the main topic on the score screen 520. When the first similarity is calculated using a reference vector instead of a main topic vector, and the score calculation unit 302 selects the first sub-score low to derive a score, the display control unit 301 may determine that the first answer is close to an answer associated with the main topic. In this case, the display control unit 301 may display on the score screen 520 a message indicating that the first answer is close to the answer associated with the topic.

[0071] When the control unit 303 determines that the score derived by the score calculation unit 302 satisfies the first condition indicating low uniqueness, the control unit 303 controls the lighting device 500 that illuminates the vicinity of the user to lower the color temperature of the illumination light and increase the illuminance.

[0072] A non-patent document (Koichiro Tomimoto et al., "Office Lighting System Aimed at Improving Worker Creativity: Research on Lighting Conditions Suitable for Creative Office Scenes," Proceedings of the Kansai Branch Conference of the Japan Ergonomics Society, Volume 2009, Pages 171-174, Published December 5, 2009) describes the use of high-intensity, low-color-temperature lighting as a suitable condition for tasks requiring creativity, such as brainstorming. Therefore, when a user is performing a task requiring divergent thinking, the control unit 303 generates a control signal that lowers the color temperature by a predetermined amount and increases the illuminance by a predetermined amount, and transmits the generated control signal to the lighting device 500 via the communication unit 320. The control unit 303 repeats the control of lowering the color temperature by a predetermined amount and increasing the illuminance by a predetermined amount until the first answer no longer satisfies the first condition. This allows a user who is not demonstrating individuality to develop their individuality.

[0073] The database device 200 includes a subject database 201, a response database 202, a conference database 203, and a user database 204. The database device 200 is configured by a computer including a processor (not shown), a memory (not shown), and a communication unit (not shown). The subject database 201 to the user database 204 are recorded in the memory.

[0074] 6 is a diagram showing the data configuration of the subject database 201 and the answer database 202. The subject database 201 records a subject ID and a subject text in association with each other. The subject ID is an identifier that uniquely identifies the subject text.

[0075] The topic text is input by a user using the input device 101. In detail, the input device 101 acquires the topic text by inputting the topic voice signal collected by the microphone 101a to a voice recognizer, and transmits the topic text to the database device 200. The database device 200 assigns a topic ID to the transmitted topic text and records it in the topic database 201. The voice recognizer may be included in the input device 101, an external server, or the database device 200. When the database device 200 has a voice recognizer, the input device 101 may transmit the topic voice signal to the database device 200.

[0076] The answer database 202 records answer texts that indicate users' answers to the topic. The answer database 202 records conference IDs, answer IDs, user IDs, answer texts, scores, similar answers, and speech times in association with each other. The conference ID is an identifier for the conference. The answer ID is an identifier for the answer text. The user ID is an identifier for the user. The answer text is text data that indicates the user's answer. The score is a score for the answer. The similar answer field records the answer ID of the past answer that is most similar to the first answer among past answers given during the same conference. The speech time is the time when the answer text was spoken.

[0077] For example, the first row of the record in the answer database 202 stores the following information: a user with user ID "1" answered "Make it a doll's head" in a conference with conference ID "1", a score of "70" was derived for that answer, there were no similar answers to that answer, and the time of speech was "January 20, 2023, 17:14:31".

[0078] An answer text is input by a user using the input device 101. In detail, the input device 101 acquires the answer text by inputting a voice signal of the answer picked up by the microphone 101a to a voice recognizer, and transmits the answer text to the database device 200. At this time, the input device 101 identifies the user who spoke the answer.

[0079] Various methods can be used to identify the user who has uttered the answer.

[0080] For example, if the microphone 101a of the input device 101 is configured as an array microphone, the input device 101 can identify which user in the booth 400 has spoken by identifying the direction of the sound source. The input device 101 prompts the user to input a seating map using the keyboard 101b and mouse 101c at the start of the conference. The seating map is information in which the seat position in the booth 400, the seat ID, and the user ID of the user sitting at the seat are associated with each other. The input device 101 can identify the user ID of the user who spoke the answer by comparing the sound source direction with the seating map. The input device 101 associates the identified user ID with the answer text and transmits them to the database device 200.

[0081] Alternatively, each of multiple users participating in the conference may carry a microphone, and the input device 101 may identify the user who spoke the answer. A user ID is associated with each microphone in advance. When the input device 101 acquires a voice signal from a certain microphone, it can identify the user who spoke the answer from the user ID associated with that microphone.

[0082] By using these methods, the database device 200 can associate the answer text with the user ID and record them in the database device 200. The database device 200 may identify the time when the answer text is received as the utterance time, or, if the answer text transmitted from the input device 101 includes the utterance time, may identify the utterance time as the utterance time of the answer text.

[0083] The database device 200 assigns the conference ID, the answer ID, and the user ID to the transmitted answer text and records it in the subject database 201.

[0084] The database device 200 may identify the conference ID of the conference in which the user is participating by comparing the user ID and speech time of the user who spoke the answer with the conference database 203, and record the conference ID in the answer database 202.

[0085] The database device 200 may acquire the score of the answer and the similar answers derived by the utterance processing device 300 from the utterance processing device 300, and record the score of the answer and the similar answers in the database device 200.

[0086] The answer database 202 records answers similar to a certain answer. Therefore, if an answer has a low score, the speech processing device 300 can determine which user's idea influenced this answer. As a result, if a first user repeatedly gives answers similar to the past answers of a second user, the first user's supervisor can determine that the first user may be overly in sync with the second user. In such a case, the supervisor can give advice to the first user, such as, "You can speak more freely."

[0087] FIG. 7 is a diagram showing the data configuration of the conference database 203 and the user database 204. The conference database 203 records information about conferences. The conference database 203 records a conference ID, a subject ID, a participant ID, a start time, and an end time in association with each other. The participant ID is the user ID of a user who participated in the conference indicated by the conference ID. The number of participant ID fields is equal to the number of users who participated in the conference. The start time is the start time of the conference indicated by the conference ID. The end time is the end time of the conference indicated by the conference ID. This associates the conference, the subject of the conference, the users who participated in the conference, the start time of the conference, and the end time of the conference. In the example of FIG. 9, one conference is associated with one subject.

[0088] When the database device 200 receives an input notification of the start conference button on the subject input screen 510 from the input device 101 at the start of the conference, it generates a conference ID. The database device 200 simply records the subject ID assigned to the subject input on the subject input screen 510, the conference ID, and the conference start time in association with each other in the conference database 203.

[0089] In the answer database 202, answer IDs are linked to conference IDs, and in the conference database 203, conference IDs are linked to subject IDs. Therefore, the database device 200 stores past texts in a storage device, linked to the subject.

[0090] The database device 200 may obtain the user IDs of all users participating in the conference from the input device 101 at the start of the conference, and record the user IDs of the users participating in the conference in the conference database 203 .

[0091] The database device 200 may identify the time when the conference end button on the score screen 520 is pressed at the end of the conference as the conference end time and record this in the conference database 203 .

[0092] The user database 204 records information about users who participated in a conference. The user database 204 records user IDs, names, affiliations, and job titles in association with each other. The name is the name of the user identified by the user ID. The affiliation indicates the department within the company to which the user identified by the user ID belongs. The job title is the user's job title within the company to which the user identified by the user ID belongs. The database device 200 acquires this information in advance and records it in the user database 204.

[0093] The lighting device 500 is installed inside the booth 400 and irradiates the interior of the booth 400 with illumination light. FIG. 8 is a diagram showing the lighting device 500 installed inside the booth 400. The booth 400 is configured as a box that separates the space where the user is located from the external space. A plurality of light sources that constitute the lighting device 500 are installed on the ceiling of the booth 400. If the score of the first answer is low, the color temperature of the lighting device 500 is lowered and the illuminance is increased. This encourages the user to create a highly original answer.

[0094] The first vector may be a word vector that has the smallest similarity to the topic vector or the reference vector among a plurality of word vectors corresponding to a plurality of words included in the first text.

[0095] The second vector may be a word vector that has the smallest similarity to the topic vector or the reference vector among a plurality of word vectors corresponding to a plurality of words included in the past text.

[0096] FIG. 9 is a diagram illustrating the calculation process of a first vector. In the example of FIG. 9, a "vacuum cleaner" is set as the topic. "Playing while straddling the handle part" is the first text D1 indicating the answer to the topic. The score calculation unit 302 separates the first text D1 using morphological analysis. As a result, the first text D1 is separated into words such as "playing while straddling the handle / part / of / the handle."

[0097] The score calculation unit 302 deletes unnecessary words, which are words with excluded parts of speech, from the segmented first text D1. As a result, "pattern," "part," "straddle," and "play" are extracted from the first text D1. The excluded parts of speech are, for example, particles and auxiliary verbs. Therefore, words with parts of speech such as nouns, adjectives, verbs, and adjectives are extracted from the first text D1.

[0098] The score calculation unit 302 calculates a word vector for each of these four words. Map 1101 shows a vector space onto which the word vectors are mapped. Of these four words, "straddle" has the smallest similarity because it is far from the thematic vector or the reference vector. Therefore, the score calculation unit 302 adopts the word vector of "straddle" as the word vector of the first vector.

[0099] In this way, by identifying the word from the first text that has the smallest similarity to the thematic vector or reference vector and using the word vector of that word to quantify the first vector, it becomes easier to evaluate long answers.

[0100] The above process is similarly applied to the second vector. Furthermore, when the topic text or the reference text contains multiple words, the score calculation unit 302 may use the average vector of the word vectors of each word as the topic vector or the reference vector. Furthermore, the score calculation unit 302 may use the average vector of each word vector of multiple words contained in the first text as the first vector, and the average vector of each word vector of multiple words contained in the second text as the second vector.

[0101] 10 is a flowchart showing an example of processing when the information processing system 1 acquires an answer. In step S101, the input device 101 detects the start of a conference. The input device 101 may determine that a conference has started when a subject is entered in the subject input field 511 on the subject input screen 510 and the start conference button 512 is pressed. If the start of the conference is detected (YES in step S101), the processing proceeds to step S102, and if the start of the conference is not detected (NO in step S101), the processing returns to step S101.

[0102] Next, in step S102, the input device 101 transmits the subject text indicating the subject entered in the subject input field 511 to the database device 200, and the database device 200 generates a subject ID corresponding to the subject text and records the subject ID and the subject text in association with each other in the subject database 201.

[0103] Next, in step S103, the database device 200 generates a conference ID corresponding to the subject ID generated in step S102, and records the subject ID, the conference ID, and the start time of the conference in the conference database 203. At this time, the database device 200 acquires, from the input device 101, the user IDs of the users who will participate in the conference input to the input device 101, and records the acquired user IDs in the conference database 203 in association with the conference ID.

[0104] Next, in step S104, the input device 101 starts accepting answers.

[0105] Next, in step S105, input device 101 detects the end of the conference. When end conference button 523 is pressed, input device 101 may determine that the conference has ended. If the end of the conference is detected (YES in step S105), the process proceeds to step S108, and if the end of the conference is not detected (NO in step S105), the process proceeds to step S106.

[0106] Next, in step S106, the input device 101 detects an answer by the user. For example, the input device 101 may detect, as the answer, a voice uttered during a period from when speech starts until a silent period is detected, or may detect, as the answer, a voice uttered by the user during a period from when an answer start button (not shown) is pressed until when an answer end button (not shown) is pressed.

[0107] If an answer is detected (YES in step S106), the process proceeds to step S107, and if an answer is not detected (NO in step S106), the process returns to step S105.

[0108] Next, in step S107, the input device 101 transmits an answer text indicating the detected answer to the database device 200, and the database device 200 generates an answer ID corresponding to this answer text, associates the answer text, the answer ID, and the speech time, and records them in the answer database 202. At this time, the input device 101 identifies the user who has answered, and transmits the user ID of the identified user to the database device 200. Therefore, the database device 200 also associates this user ID with the answer text and records it in the answer database 202.

[0109] When the process of step S107 ends, the process returns to step S105.

[0110] In step S108, the input device 101 transmits a conference end instruction to the database device 200, and the database device 200 records the conference end time in the conference database 203. When the process of step S108 ends, the process of the flowchart ends.

[0111] That is, once the conference starts, each time a user makes a response, the response text of that response is accumulated in the response database 202 .

[0112] 11 is a flowchart showing an example of processing when the information processing system 1 evaluates an answer. In step S201, the score calculation unit 302 determines whether the conference start time has been recorded in the conference database 203. Note that in the above-mentioned step S103, if the database device 200 has recorded the conference start time in the conference database 203, it transmits a conference start notice to the speech processing device 300. Therefore, if the conference start notice is received by the speech processing device 300, the score calculation unit 302 may determine that the conference start time has been recorded in the conference database 203. If it is determined that the conference start time has been recorded in the conference database 203 (YES in step S201), the processing proceeds to step S202, and if it is not determined that the conference start time has been recorded in the conference database 203 (NO in step S201), the processing returns to step S201.

[0113] Next, in step S202, the score calculation unit 302 determines whether the conference end time has been recorded in the conference database 203. In the above-mentioned step S108, if the database device 200 has recorded the conference end time in the conference database 203, it transmits a conference end notice to the speech processing device 300. Therefore, if the conference end notice is received by the speech processing device 300, the score calculation unit 302 may determine that the conference end time has been recorded in the conference database 203. If it is determined that the conference end time has been recorded in the conference database 203 (YES in step S202), the process ends, and if it is not determined that the conference end time has been recorded in the conference database 203 (NO in step S202), the process proceeds to step S203.

[0114] Next, in step S203, the score calculation unit 302 determines whether or not a new answer text has been added to the answer database 202. Note that in the above-mentioned step S107, when the database device 200 adds a new answer text to the answer database 202, it transmits an answer addition notification to the utterance processing device 300. When the answer addition notification is received by the utterance processing device 300, the score calculation unit 302 may determine that a new answer has been added. This new answer text is the first text, and the answer indicated by this new answer text is the first answer. When it is determined that a new answer text has been recorded in the answer database 202 (YES in step S203), the process proceeds to step S204, and when it is not determined that a new answer text has been recorded in the answer database 202 (NO in step S203), the process returns to step S203.

[0115] Next, in step S204, the score calculation unit 302 inputs the topic text and the first text into a language model to obtain a topic vector and a first vector, and derives a first similarity between the topic vector and the first vector. Note that the first similarity may be derived using a reference vector instead of the topic vector.

[0116] Next, in step S205, the score calculation unit 302 derives a first sub-score from the first similarity.

[0117] Next, in step S206, the score calculation unit 302 derives a second similarity between the first vector and the second vector. The score calculation unit 302 may acquire past response texts of the currently ongoing conference as past texts from the response database 202 and input the acquired past texts into a language model to acquire the second vector. When multiple past texts are acquired, the score calculation unit 302 derives the second similarity for each of the multiple past texts. When the past texts are not recorded in the response database 202 immediately after the start of the conference, the score calculation unit 302 may derive the score of the first response as 0. Note that the past texts used to calculate the second vector may be response texts of previously held conferences that have the same subject as the currently ongoing conference, in addition to (or instead of) the past texts of the currently ongoing conference.

[0118] Next, in step S207, the score calculation unit 302 derives a second sub-score from the second similarity. If multiple second similarities are derived, the score calculation unit 302 may derive a second sub-score for each of the multiple second similarities.

[0119] Next, in step S208, the score calculation unit 302 selects the first subscore and the second subscore in descending order to derive the score of the first answer. If there are multiple second subscores, the score calculation unit 302 may select the smallest subscore from the multiple second subscores and the first subscore, and derive the selected smallest subscore as the score of the first answer.

[0120] Next, in step S209, the display control unit 301 displays the score of the first answer on the display device 102. In this case, the display control unit 301 may display the first answer and the score of the first answer on the score screen 520.

[0121] 12 is an explanatory diagram of the process of deriving a score. In this example, the user answers how to use the topic "ping pong." The answer IDs are assigned numbers according to the order of the answers.

[0122] The "Max word" refers to the word that has the maximum subscore obtained from the similarity to the topic among the words that make up the answer text.

[0123] The first subscore indicates the subscore of the Max word and the topic. The answer history includes all Max words at the time of the answer. For example, the answer history field of the record with answer ID "3" records "doll" and "god." This is because the Max word in the answer text for answer ID "1" was "doll," and the Max word in the answer text for answer ID "2" was "god."

[0124] The Min word is the word with the smallest subscore obtained from the similarity with the Max word among the words listed in the answer history field. For example, the Min word for the record with answer ID "3" is "doll." This is because, among the words "doll" and "god" listed in the answer history field for answer ID "3," the word with the smallest subscore with the Max word "snake" is "doll."

[0125] The second subscore is the subscore of the Max word and the Min word.

[0126] The score is the lower of the first and second subscores. In the score field, underlined scores indicate scores where the second subscore is selected, and non-underlined scores indicate scores where the first subscore is selected.

[0127] In process (a), the score calculation unit 302 derives a first sub-score based on the similarity between the new answer (first answer) and the topic. In the example of answer ID "8," unnecessary words are removed from the answer "Punch holes to make earrings," and "holes," "pierce," and "earrings" are extracted. Three similarities between the word vectors of these three words and the topic vector are derived, and three sub-scores are derived from the three derived similarities. As a result of the derivation, "earrings" has the highest sub-score of the three words, so "earrings" is determined as the Max word. The sub-score of "earrings," "1.007," is determined as the first sub-score.

[0128] In step (b), the score calculation unit 302 derives a second sub-score based on the similarity between the Max word and each word recorded in the answer history field. In the example of answer ID "8," seven similarities between "earring" and seven words, "doll," "god," "snake," "necklace," "plate," "slope," and "health," are derived, and seven sub-scores are derived from the seven similarities. The word "necklace," which corresponds to the smallest sub-score of the seven sub-scores, is determined as the Min word, and the sub-score of the Min word is determined as the second sub-score.

[0129] In step (c), the score calculation unit 302 selects the lower of the first sub-score and the second sub-score to derive a final score. In the example of the record with answer ID "8", the lower of the first sub-score "1.007" and the second sub-score "0.131", the second sub-score "0.131", is derived as the score for the answer with answer ID "8".

[0130] In step (d), the Max word is recorded in the answer history field. In the example of answer ID "8", the Max word "earrings" is added to the answer history field of the record with answer ID "8".

[0131] As such, the answer to the ping pong game "punch holes in them and turn them into earrings" has a very high subscore with the topic of "1.007." However, because a similar answer, "punch holes in them and turn them into necklaces," has been given in the past, the answer "punch holes in them and turn them into earrings" has a small final score of "0.131." Therefore, even if an answer has a high subscore with the topic, the score calculation unit 302 can derive a low score if the answer is similar to a past answer. As a result, the score calculation unit 302 can accurately determine the uniqueness of the answer.

[0132] In this way, the information processing system 1 derives a score based on the most similar topic or past answers. Therefore, the information processing system 1 can more accurately inform meeting participants of the good or bad score. Furthermore, the information processing system 1 also presents the reasons for deriving the score, allowing the user to come up with a more original answer.

[0133] (Variation) (1) Although the speech processing device 300 derives the thematic vector and the first vector from the thematic text and the first text, the present disclosure is not limited to this, and the input device 101 may derive the thematic vector and the first vector. In this case, the database device 200 may record the thematic vector acquired from the input device 101 in the thematic database 201, and record the first vector acquired from the input device 101 in the answer database 202. The speech processing device 300 may acquire the thematic vector from the thematic database 201, and the first vector from the answer database 202. Furthermore, in this case, since the first vector is recorded in the answer database 202, the second vector is also recorded in the answer database 202. Therefore, the speech processing device 300 can acquire the second vector from the answer database 202.

[0134] (2) The conversion from the speech signal to text may be performed by the speech processing device 300. In this case, the database device 200 may record the speech signal of the topic in the topic database 201 and the speech signal of the answer in the answer database 202. The speech processing device 300 may acquire the speech signals of the topic and the answer from the database device 200, input the speech signals of the topic and the answer to a speech recognizer that is owned by the speech processing device 300 or that is owned by an external server, and acquire the topic text and the first text.

[0135] (3) The information processing system 1 may be configured with the speech processing device 300, or may be configured with the database device 200 and the speech processing device 300. [Industrial Applicability]

[0136] The present disclosure is useful for generating user ideas. [Explanation of symbols]

[0137] 1: Information processing system 101: Input device 102:Display device 200: Database device 201: Subject Database 202: Response database 203: Conference Database 204: User database 300: Speech processing device 301: Display control unit 302: Score calculation unit 303: Control section 310: Processor 320: Communications Department 330: Memory

Claims

1. An information processing method in a computer, comprising: obtaining a first vector corresponding to a first text indicating a first answer to the topic; Deriving a first sub-score corresponding to the level of uniqueness based on a first similarity indicating the similarity between the first vector and a topic vector corresponding to topic text indicating the topic, or the similarity between the first vector and a reference vector corresponding to reference text indicating an answer associated with the topic; obtaining a second vector corresponding to a past text indicating a response made to the topic earlier than the first response; deriving a second sub-score based on a second similarity indicating a similarity between the first vector and the second vector; deriving a score indicating a degree of originality of the first answer based on a lower subscore of the first subscore and the second subscore; Information processing methods.

2. the first similarity and the second similarity each correspond to a cosine similarity; The information processing method according to claim 1 .

3. and if the score satisfies a first condition indicating low uniqueness, notifying the user who provided the first answer of the cause of the low score.

3. The information processing method according to claim 1 or 2.

4. The notification of the cause includes notifying that the first answer is close to a past answer of a specific person. The information processing method according to claim 3 .

5. The notification of cause includes notifying the first answer that it is not original. The information processing method according to claim 3 .

6. The notification of the cause includes notifying that the first answer is close to any of a past answer, the subject, and an answer associated with the subject. The information processing method according to claim 3 .

7. When the score satisfies a first condition indicating that the uniqueness is low, a lighting device that illuminates the vicinity of the respondent is controlled to lower the color temperature of the lighting light and increase the illuminance.

3. The information processing method according to claim 1 or 2.

8. The past text is stored in a storage device in association with the subject.

3. The information processing method according to claim 1 or 2.

9. the first vector is a word vector that has the smallest similarity to the subject vector or the reference vector among a plurality of word vectors corresponding to a plurality of words included in the first text, the second vector is a word vector having the smallest similarity to the subject vector or the reference vector among a plurality of word vectors corresponding to a plurality of words included in the past text; 3. The information processing method according to claim 1 or 2.

10. An information processing system including a computer, The computer obtaining a first vector corresponding to a first text indicating a first answer to the topic; deriving a first sub-score corresponding to the level of uniqueness based on a first similarity indicating the similarity between the first vector and a topic vector corresponding to topic text indicating the topic, or the similarity between the first vector and a reference vector corresponding to reference text indicating an answer associated with the topic; obtaining a second vector corresponding to a past text indicating an answer given to the topic earlier than the first answer; deriving a second sub-score based on a second similarity indicating a similarity between the first vector and the second vector; a processor that executes processing to derive a score that indicates a degree of originality of the first answer based on the lower subscore of the first subscore and the second subscore; Information processing system.

Images