Information processing device, presentation method, and presentation program

Abstract

An information processing device (10) includes: a generation unit (15A) for generating, in accordance with a frame of reference implicitly used when a human verbalizes a complex concept, an element serving as a clue to the meaning of a word used for linguistic communication; and a presentation unit (15B) for associating each frame of reference with a corresponding element and presenting the same.

Description

Information processing device, presentation method, and presentation program

[0001] This invention relates to an information processing device, a presentation method, and a presentation program.

[0002] The meanings of words used in linguistic communication can be broadly categorized into objective meanings, which are independent of the speaker, and subjective meanings, which are dependent on the speaker. Of these, subjective meanings are prone to misunderstandings among participants and are a major cause of miscommunication.

[0003] Furthermore, it has become clear that even highly subjective language, sometimes referred to as "emotional language," can improve the performance of highly skilled groups if communicated accurately. In other words, accurately conveying subjective meanings to the recipient is crucial for enhancing mutual understanding and creativity.

[0004] For example, one technology proposed to support language communication is a conventional technique that visualizes individual differences in impressions of words by mapping the impressions of meeting participants on the topic onto a two-dimensional space.

[0005] Masao Ohira, et al. EVIDII: A system to support mutual understanding through the visualization of differences. Transactions of the Information Processing Society of Japan, 2000, 41.10: 2814-2826.

[0006] However, the clues mapped onto a two-dimensional space using the conventional techniques described above are based on pre-prepared, fixed datasets, such as words or images. This limits the granularity of semantic decomposition, thus restricting the application scenarios and leaving room for improvement.

[0007] Therefore, the present invention aims to provide an information processing device, a presentation method, and a presentation program that can expand the range of application scenarios in which language communication can be supported.

[0008] To solve the above-mentioned problems and achieve the objective, the information processing device of the present invention comprises a generation unit that generates elements that serve as clues to the meaning of words used in linguistic communication, along axes implicitly used by humans when verbalizing complex concepts, and a presentation unit that associates and presents elements corresponding to each axis.

[0009] According to the present invention, the range of application scenarios in which language communication can be supported can be expanded.

[0010] Figure 1 is a block diagram (1) showing an example of the functional configuration of an information processing device. Figure 2 is a schematic diagram showing an example of the visualization of differences using conventional technology. Figure 3 is a diagram showing one aspect of the problem-solving approach. Figure 4 is a schematic diagram explaining one aspect of the effect. Figure 5 is a diagram showing an example of a prompt. Figure 6 is a diagram showing an example of a presentation screen. Figure 7 is a diagram showing an example of a selection result screen. Figure 8 is a flowchart showing the procedure for presentation processing. Figure 9 is a diagram showing an example of the processing content in application scene 1. Figure 10 is a diagram showing an example of a selection result screen in application scene 1. Figure 11 is a diagram showing an example of the processing content in application scene 2. Figure 12 is a diagram showing an example of a selection result screen in application scene 2. Figure 13 is a flowchart showing the procedure for presentation processing related to an application example. Figure 14 is a diagram showing an example of the processing content in application scene 3. Figure 15 is a diagram showing an example of a selection result screen in application scene 3. Figure 16 is a block diagram (2) showing an example of the functional configuration of an information processing device. Figure 17 is a diagram showing an example of a prompt. Figure 18 is a diagram showing an example of a meeting screen. Figure 19 is a flowchart showing the procedure for affective information extraction processing. Figure 20 shows an example of a prompt. Figure 21 is a block diagram (3) showing an example of the functional configuration of the information processing device. Figure 22 shows an example of a usage history DB. Figure 23 is a flowchart showing the procedure for the usage history extraction process. Figure 24 shows an example of the processing content for application scene 4. Figure 25 shows an example of the selection result screen in application scene 4. Figure 26 is a block diagram (4) showing an example of the functional configuration of the information processing device. Figure 27 is a flowchart showing the procedure for the axis update process. Figure 28 shows an example of the hardware configuration.

[0011] The following description will explain the embodiments for implementing the information processing device, presentation method, and presentation program related to this disclosure (hereinafter referred to as "Embodiments") with reference to the attached drawings. It should be noted that these embodiments represent only one example or aspect, and the following description does not limit the structure, operation, function, properties, characteristics, methods, and applications related to this disclosure.

[0012] <Embodiment 1> <Overall Configuration> Figure 1 is a block diagram (1) showing an example of the functional configuration of the information processing device 10. For example, Figure 1 shows an information processing device 10 that provides a presentation function to generate and present elements that serve as clues to the subjective meaning of words used in linguistic communication, along axes implicitly used by humans when verbalizing complex concepts.

[0013] The following is merely one example of a usage scenario in which language communication takes place, using conferencing system 3, which provides communication functions for sharing video, audio, documents, etc., between multiple locations.

[0014] In one embodiment, the information processing device 10 may be implemented by a server device. For example, the information processing device 10 can provide the above-mentioned functions as a cloud service by running a PaaS (Platform as a Service) type middleware or a SaaS (Software as a Service) type application.

[0015] As shown in Figure 1, the information processing device 10 can be connected to the conference system 3 via a network NW in a communicative manner. For example, the network NW may be implemented by any type of communication network, such as the Internet or a LAN (Local Area Network), whether wired or wireless.

[0016] Conference System 3 is a system that provides communication functions for sharing video, audio, documents, and other materials between multiple locations. For example, Conference System 3 may be implemented using any type of web conferencing system, such as on-premise, cloud-based, or browser-based. However, it is not limited to these, and Conference System 3 may also be implemented using a video conferencing system.

[0017] Hereafter, users participating in a meeting provided by the communication function of conferencing system 3 may be referred to as "participants." Furthermore, among the participants, users who initiate communication may be referred to as "senders," while users who receive communication may be referred to as "receivers."

[0018] While the above-mentioned functionality is presented here as an example of being provided as a cloud service, it is not limited to this. For example, the above-mentioned functionality may be provided on-premises. Furthermore, the above-mentioned functionality may be packaged as a feature of a service or application provided by a service provider offering services related to remote conferencing.

[0019] Furthermore, while the above presentation function is implemented as an example in a client-server system, it is not limited to this. For example, the above presentation function may be provided as a standalone function by having an application running on a user terminal participating in a meeting conducted by conference system 3 execute processing corresponding to the above presentation function on the user terminal.

[0020] <One aspect of the challenge> As explained in the background technology section above, the clues mapped onto the two-dimensional space by the conventional technology described above are pre-prepared, fixed datasets, such as words and images. Since there is a limit to the granularity of semantics that can be decomposed, there is room for improvement in that the application scenarios are limited.

[0021] Figure 2 is a schematic diagram illustrating an example of visualizing differences using conventional technology. As an example, Figure 2 shows an example in which Hanako and Taro are asked to select their impressions of the word "refreshing" from a pre-prepared image dataset. As a result, by comparing the images selected by Hanako and the images selected by Taro, the individual differences in the subjective meaning of "refreshing" between Hanako and Taro can be visualized.

[0022] However, with the conventional techniques described above, the granularity at which the meaning of words can be broken down is limited to the granularity of the images registered in the dataset. Therefore, if an image matching the user's impression is not registered in the dataset, it is difficult to visualize individual differences in the subjective meaning of the same word among users, and the application scenarios are limited.

[0023] <One aspect of the problem-solving approach> Therefore, in this embodiment, a presentation function is provided that dynamically generates and presents elements that serve as clues to the subjective meaning of words used in linguistic communication, along the axis that humans implicitly use when verbalizing complex concepts.

[0024] Figure 3 illustrates one aspect of a problem-solving approach. Figure 3 shows an example of using three-dimensional axes—an "abstract axis," a "time axis," and a "spatial axis"—for segmentation, as an example of axes implicitly used when humans verbalize complex concepts. Furthermore, Figure 3 shows an example of how three elements—"end of last year," "strong citrus scent," and "abruptness"—are dynamically generated along the abstract, time, and spatial axes, as an example of semantic clues to the word "refreshing." For example, in the example shown in Figure 3, the element "end of last year" is presented in relation to the spatial axis. Additionally, the element "strong citrus scent" is presented in relation to the intersection of the abstract and time axes, while the element "abruptness" is presented in relation to the intersection of the time and spatial axes. This makes it possible to share the subjective meaning of words used in linguistic communication deeply and accurately among participants to a degree that is valuable to them.

[0025] Figure 4 is a schematic diagram illustrating one aspect of the effect. Figure 4 schematically illustrates the difference in the content presented as verbal cues between the prior art and this embodiment. As shown in Figure 4, the prior art lacks appropriate cues for semantic decomposition, especially for words with strong subjectivity. That is, the prior art uses a pre-prepared fixed dataset, which inevitably results in static and discrete cues. As a result, only cues e1 and e2, which have little relevance to the concept 20 that the sender wants to express, can be selected, and the difference between the concept 20 that the sender wants to express and its similar concepts remains unclear. On the other hand, the presentation function according to this embodiment can present dynamically and continuously generated cues along axes A1 to A4, which humans implicitly use when verbalizing complex concepts, even for words with strong subjectivity. As a result, cues with a high relevance to the concept 20 that the sender wants to express are presented, and the difference between the concept 20 that the sender wants to express and its similar concepts can be clarified.

[0026] Thus, the presentation function according to this embodiment dynamically and continuously generates semantic cues for subjective words used in linguistic communication, and the granularity of semantics that can be decomposed can also be dynamically changed.

[0027] Therefore, the presentation function according to this embodiment can expand the range of application scenarios in which it can support verbal communication. Furthermore, since the presentation function according to this embodiment can accurately convey even the subtle elements of the subjective meaning of words used in verbal communication, the accuracy of cues can also be improved.

[0028] <Configuration of Information Processing Device 10> Next, the functional configuration of the information processing device 10 that provides the above-mentioned presentation function will be described. Figure 1 schematically shows the blocks related to the presentation function of the information processing device 10. As shown in Figure 1, the information processing device 10 has a communication control unit 11, a storage unit 13, and a control unit 15. Note that Figure 1 only shows a selection of the functional units related to the above-mentioned presentation function, and the information processing device 10 may also be equipped with functional units other than those shown.

[0029] The communication control unit 11 is a functional unit that controls communication with other devices such as the conference system 3. In one embodiment, the communication control unit 11 can be implemented by a network interface card such as a LAN card. In one aspect, the communication control unit 11 receives speech data or text data of utterances input via voice input during a conference conducted by the conference system 3, or outputs information presented by the above-mentioned presentation function to user terminals participating in the conference provided by the conference system 3.

[0030] The storage unit 13 is a functional unit that stores various types of data. In one embodiment, the storage unit 13 may be implemented by internal, external, or auxiliary storage of the information processing device 10. For example, the storage unit 13 stores model information 13A. The model information 13A will be explained later in conjunction with the scenes in which each type of data is referenced or registered.

[0031] The control unit 15 is a functional unit that performs overall control of the information processing device 10. For example, the control unit 15 can be implemented by a hardware processor. As shown in Figure 1, the control unit 15 has an element generation unit 15A and an element presentation unit 15B. The control unit 15 may also be implemented by hardwired logic or the like.

[0032] The element generation unit 15A has the function of generating elements that serve as clues to the meaning of words used in linguistic communication. Here, "words" may be single words, phrases containing multiple words, or sentences.

[0033] In one embodiment, the element generation unit 15A acquires information (hereinafter referred to as "symbols") that expresses the subjective feelings of the speaker among the participants in a meeting conducted by the conference system 3. As merely one example of such symbols, it is possible to perform semantic segmentation targeting words that are particularly subjective, known as "emotional words." For example, the element generation unit 15A can acquire emotional words from the speech text corresponding to the utterances input into the conference system 3. Then, the element generation unit 15A inputs a prompt with the embedded symbol into an arbitrary generative model or generative AI (Artificial Intelligence), causing the LLM 5 to generate elements that serve as clues to the meaning of the symbol.

[0034] Figure 5 shows an example of a prompt 50. As shown in Figure 5, the prompt 50 includes an instruction sentence that instructs a large language model, or LLM (Large Language Model) 5, to perform the task of decomposing the meaning of an affective word (X) along an axis.

[0035] Such an LLM5 may be executed in the information processing device 10 based on model information 13A stored in the memory unit 13. For example, the model information 13A may include hyperparameters such as the layer structure of the LLM5, and parameters such as the weights and biases of the LLM5. Note that the LLM5 does not necessarily have to be executed in the information processing device 10; it is also possible to call a process that causes the LLM5 to execute prompt 50 by sending an API (Application Programming Interface) to an MLaaS (Machine Learning as a Service) or the like that provides the functionality of the LLM5.

[0036] Furthermore, as shown in Figure 5, the prompt 50 is embedded with the value X of the affective word to be analyzed in the instruction sentence, as well as an example of the segmentation axis to be used during the analysis.

[0037] For example, in the example shown in FIG. 5, examples include an axis A defined by an abstractness level, an axis B defined by human physical sensations, an axis C defined by spatial positional relationships, an axis D defined by time series, and an axis E defined by purposes and functions.

[0038] Among these axes A to E, for axis A and axes C to D, a one-dimensional axis can be defined by binary opposition. For example, in the case of axis A, the abstractness level is defined by the degrees of abstraction and concreteness. Further, in the case of axis C, space is defined by distance and proximity. Furthermore, in the case of axis D, time is defined by the past and the future. On the other hand, in the case of axis B, the degree of relevance to each sensory organ such as vision, hearing, smell, taste, and touch can be defined in multiple dimensions. Furthermore, in the case of axis E, the degree of relevance for each type of purpose or function can be defined in multiple dimensions.

[0039] The LLM 5 to which such a prompt 50 is input outputs elements that serve as clues to the meaning indicated by the sensuous word X. Furthermore, by embedding an instruction to generate clues stepwise along the axes together with the definitions of axes A to E, continuously generated clues on each axis can be obtained.

[0040] Here, as an example of a clue, an example of generating text in which clues to the meaning of a sensuous word are described in natural language is given, but data of other modalities, such as images or audio, can also be generated. Furthermore, although a sensuous word is given as an example of a symbol, information that can be sensed when the sensuous word is uttered, such as intonation, sound pressure, the amount of silence, facial expression, posture, gesture, body odor, physiological phenomena, physical condition, and the like, may be acquired as symbols. Furthermore, although an example using a generative model is given as an example of an element generation method, a generative model does not necessarily have to be used; a rule base describing rules for generating elements may be used, or a machine learning model may be used.

[0041] The element presenting unit 15B has a function of presenting elements generated for each axis by the element generating unit 15A. Fig. 6 is a diagram showing an example of a presentation screen 200. As shown in Fig. 6, on the presentation screen 200, elements corresponding to respective axes such as an "abstract axis", a "sensory axis", a "space axis", a "time axis" and a "purpose" are displayed in association with each axis. For example, for each axis, text describing the elements generated by the element generating unit 15 in natural language can be mapped and displayed. Such a presentation screen 200 may be presented, merely as an example, to user terminals of users such as participants attending a conference implemented by the conference system 3 and persons related thereto. At this time, when a plurality of elements are generated by the element generating unit 15, the display form of the elements can also be changed according to the importance of the elements. For example, when the element is text, the size of the text display area and the degree of font emphasis are set to be higher as an index for measuring the information amount of text, for example, TF-IDF (Term Frequency-Inverse Document Frequency), increases, whereby an element having a larger index can be displayed more conspicuously.

[0042] It should be noted that although Fig. 6 shows an example in which one or more elements are displayed on each axis, elements do not necessarily have to be generated for all axes. In addition, although Fig. 6 shows an example in which a plurality of elements are displayed on one axis, a plurality of elements do not necessarily have to be plotted. Furthermore, although Fig. 6 shows an example in which the presentation of elements is realized by display output, the presentation content can also be printed out, or synthesized voice reading out the presentation content can be output as voice.

[0043] As a further aspect, the element presentation unit 15B can also accept the selection of one or more elements from among multiple elements. Figure 7 shows an example of the selection result screen 210. As shown in Figure 7, the selection result screen 210 displays an excerpt of the elements selected from the presentation screen 200 shown in Figure 6, and these excerpted elements are marked with black circles. For example, in the example shown in Figure 7, an example is shown in which the element mapped to the third position from the top of the abstract axis, the element mapped to the second position from the top of the sensory axis, and the element mapped to the fourth position from the top of the spatial axis were selected from the elements included in the presentation screen 200 shown in Figure 6. Such element selections are merely examples and can be accepted from the symbol sender, receiver, or even third-party user terminals other than participants. In this case, the elements do not necessarily have to be selected with a single weight. For example, the weight of the element can be specified when the element is selected. In this way, when an element is selected on the presentation screen 200, the selection result of that element can be presented as the selection result screen 210. Hereinafter, the operation of selecting elements generated by the element generation unit 15A or assigning weights to selected elements may be referred to as "annotation".

[0044] <Processing Flow> Figure 8 is a flowchart showing the steps of the presentation process. As shown in Figure 8, the element generation unit 15A acquires symbols that represent the subjective mental image of the presenter among the participants of the meeting conducted in the conference system 3 (step S101).

[0045] Next, the element generation unit 15A inputs a prompt containing the symbol acquired in step S101 to the LLM5, causing the LLM5 to generate elements that serve as clues to the meaning of the symbol along the axis (step S102).

[0046] Then, the element presentation unit 15B presents the elements generated for each axis in step S102 (step S103). After that, the element presentation unit 15B accepts a selection from the elements presented in step S103 (step S104). Finally, the element presentation unit 15B presents the selection result of the elements accepted in step S104 (step S105), and the process ends.

[0047] <Summary of Embodiment 1> As described above, the information processing device 10 according to this embodiment dynamically generates and presents elements that serve as clues to the subjective meaning of words used in linguistic communication, along an axis implicitly used by humans when verbalizing complex concepts. Therefore, the granularity of the semantic components that can be decomposed can also be dynamically changed. Accordingly, the information processing device 10 according to this embodiment can expand the range of application scenes in which it can support linguistic communication. Furthermore, the information processing device 10 according to this embodiment can accurately convey even the subtle elements of the subjective meaning of words used in linguistic communication, thus improving the accuracy of the clues.

[0048] <Application Examples> Next, application scenarios 1 and 2 will be given as examples of how the information processing device 10 according to this embodiment can be applied.

[0049] (1) Application Scenario 1 As the first example, we will consider Application Scenario 1, which involves supporting the formulation of requirements definitions in a dialogue between an IT consultant and a client. Such Application Scenario 1 may have the following background. For example, client A has an image of the service they want to realize, but they don't know the appropriate words to express it. So, from the words that client A knows, they happen to choose "time machine" as the closest. However, consultant B advises them, "No, no, let's be a little more realistic," and client A falls silent. We want to help client A share the meaning they want to convey with the word "time machine." The information processing device 10 is applied under these circumstances.

[0050] Figure 9 shows an example of the processing content in application scene 1. As shown in Figure 9, "I want to realize a service like a time machine" is input as an example of text. This text is part of a conversation that took place in a scene where client A communicates the image of the service that client A wants to realize to IT consultant B. Based on this input, the element generation unit 15A extracts the sentiment word X "time machine" from the text "I want to realize a service like a time machine" (step S1). Next, the element generation unit 15A decomposes the meaning of the sentiment word X "time machine" along multiple axes (step S2). Then, the element presentation unit 15B receives annotations for the elements that have been mapped to each axis and presented via client A's user terminal (step S3). As a result, the selection result screen 210 shown in Figure 10 is presented to the participant's user terminal.

[0051] Figure 10 shows an example of the selection result screen 210 in application scenario 1. Figure 10 shows an excerpt of elements mapped to the objective axis from among multiple axes. As shown in Figure 10, among the elements arranged on the objective axis in the order of "acquiring experiences that transcend time," "backing up past experiences," and "physical travel to the future or past," moving from concept to physical, client A has selected "acquiring experiences that transcend time." By sharing such selection results with participants, including consultant B, the following effects can be obtained in application scenario 1. For example, the clue "acquiring experiences that transcend time" presented by the element presentation unit 15B can convey that what client A actually wanted to communicate was not "physical travel to the future or past," but "an experience in which the emotions of the people at that time vividly come to life." This can support the formulation of the service concept in application scenario 1, where requirements definition is formulated in the dialogue between the IT consultant and the client.

[0052] (2) Application Scenario 2 As a second example, we will consider Application Scenario 2, which involves supporting the sharing of complex concepts in a dialogue between researchers brainstorming ideas. Such Application Scenario 2 may have the following background: For example, Researcher A and Researcher B are discussing a new technology. Researcher A has an image of the technology he wants to realize, but he doesn't know the right words to express it. So, he chooses the phrase that he thought was the closest to it from the words he could come up with: "a technology that understands humans as if you were stroking them." Researcher B replies, "What do you mean? Are you going to copy them with a 3D printer?" Researcher A knew that there was a concept that he couldn't fully express with those words, but he couldn't explain what it was, and the discussion stalled. The information processing device 10 is applied under these circumstances.

[0053] Figure 11 shows an example of the processing content in application scene 2. As shown in Figure 11, "technology to understand by caressing the object" is input as an example of text. This text is part of a dialogue that took place in a scene where researcher A communicates to researcher B the image of a new technology that researcher A wants to realize. Based on this input, the element generation unit 15A extracts the emotional word X "caress" from the text "technology to understand by caressing the object" (step S1). Next, the element generation unit 15A decomposes the meaning of the emotional word X "caress" along multiple axes (step S2). Then, the element presentation unit 15B receives annotations for the elements that have been mapped to each axis and presented via researcher A's user terminal (step S3). As a result, the selection result screen 210 shown in Figure 12 is presented to the participant's user terminal.

[0054] Figure 12 shows an example of the selection result screen 210 in application scene 2. Figure 12 shows excerpts of elements mapped to the abstract axis and the sensory axis among multiple axes. As shown in Figure 12, elements are mapped to the abstract axis in the order of "multi-faceted and all-encompassing understanding," "thorough and complete," and "dispatch one's little self around the object," moving from abstract to concrete. Furthermore, elements are mapped to the sensory axis in the order of "slid in," "multi-faceted observation," "nuzzle," "listen," "lick," and "encircle," moving from head to toe. In this way, among the elements mapped to the abstract axis, "dispatch one's little self around the object" is annotated by Researcher A, and among the elements mapped to the sensory axis, "slid in" is annotated by Researcher A. By sharing these annotation results with participants, including Researcher B, the following effects can be obtained in application scene 2. For example, the clues presented by the element presentation unit 15B allow for the precise sharing of researcher A's personal image of "experiencing the object with one's whole body" and "delving into the inner world of the object." This enables the sharing of new technology concepts in application scenario 2, where complex concepts are supported in idea generation dialogues among researchers.

[0055] <Application Example of Embodiment 1> In Embodiment 1 described above, an example was given in which elements obtained by instructing LLM5 to decompose the meaning of a symbol are presented. However, it is also possible to instruct LLM5 to recursively decompose the meaning of an element.

[0056] Figure 13 is a flowchart showing the steps of the presentation process related to the application example. Note that in Figure 13, steps that perform processes different from those performed in the flowchart shown in Figure 8 are assigned different step numbers.

[0057] For example, the flowchart shown in Figure 13 differs from the flowchart shown in Figure 8 in that the process in step S111 is added after the process in step S103.

[0058] In other words, after step S103 is executed, the element generation unit 15A determines whether or not to re-decompose one or more of the elements generated for each axis (step S111). The necessity of such re-decomposition can be determined, as an example, by whether or not a request is received from the user terminal of the sender who sends the words corresponding to the symbol, the user terminal of the receiver, or the user terminal of a third party other than the participant, specifying the elements for which re-decomposition should be performed. In addition to user requests, it is also possible to register in advance a setting to automatically perform a predetermined number of re-decompositions.

[0059] If it is determined that the element needs to be re-decomposed (step S111: Yes), the element generation unit 15A inputs a prompt to the LLM 5 that includes the element that was determined to need to be re-decomposed, causing the LLM 5 to regenerate elements that serve as clues to the meaning of the element along the axis (step S102). If it is determined that the element needs to be re-decomposed (step S111: No), the process proceeds to step S104.

[0060] By adding the process in step S111, depending on whether or not there is a request from the sender, the element generation unit 15A can perform a recursive decomposition of the meaning of the elements until the concept 20 that the sender wants to express is generated.

[0061] As an example of an application where the information processing device 10 relating to such an application example is applied, we will illustrate with application scenario 3.

[0062] (3) Application Scenario 3 As a third example, we will consider Application Scenario 3, which involves providing expressive support to improve the performance of collaborative work. Such Application Scenario 3 may have the following background. For example, in a dialogue between driver A and engineer B in motorsports, where they are considering machine settings to improve lap times, they want to evaluate how much of driver A's intuition is included in driver A's responses to engineer B's questions. The information processing device 10 relating to the application example is applied under this background.

[0063] Figure 14 shows an example of the processing content for application scene 3. As shown in Figure 14, "Hmm, it feels a bit weak" is input as an example of text. This text is part of a conversation that took place in a scene where driver A communicates to engineer B his impressions of the machine's behavior during a test run on a circuit. Based on this input, the element generation unit 15A extracts the emotion word X "weak" from the text "Hmm, it feels a bit weak" (step S1). Next, the element generation unit 15A decomposes the meaning of the emotion word X "weak" along multiple axes (step S2). Then, the element presentation unit 15B receives annotations for the element "abruptness" mapped to the abstract axis via driver A's user terminal (step S3). Furthermore, the element generation unit 15A receives an instruction from engineer B's user terminal to re-decompose the element "abruptness" along the spatial axis, and the element presentation unit 15B receives an annotation from driver A's user terminal for the element "moving to the left" which has been mapped onto the spatial axis by the re-decomposition (step S4). As a result, the selection result screen 210 shown in Figure 15 is presented to the participant's user terminal.

[0064] Figure 15 shows an example of the selection result screen 210 in application scene 3. As shown in Figure 15, during the first decomposition, the selection result screen 211 is displayed, in which driver A has annotated "abruptness" among the elements arranged on the abstract axis in the order of "surprise," "dissatisfaction," "strong impact," "abruptness," "strong impact near the pelvis," and "physical experience of the hips collapsing." Subsequently, on the selection result screen 211, engineer B gives an instruction to further decompose the element "abruptness" on the spatial axis, and the second decomposition is performed. That is, during the second decomposition, the selection result screen 212 is further displayed, in which driver A has annotated "progressing to the left" among the elements arranged on the spatial axis in the order of "upward movement," "up and down movement," "progressing to the left," and "progressing to the right." By sharing such selection results with driver A and engineer B, the following effects can be obtained in application scene 3. For example, from driver A's perspective, the elements and their search (element presentation unit 15B) can help them notice any excess or deficiency in the meaning they wanted to convey through the "collapsed tic." On the other hand, engineer B can use the elements and their search (element presentation unit 15B) as a starting point for further exploration of the "collapsed tic." This allows for improved collaborative work performance in application scenario 3, where expressive support is provided to improve collaborative work performance.

[0065] <Embodiment 2> Next, an example of the functional configuration of the information processing device 20 according to Embodiment 2 will be described. In this embodiment, a variation of the method for extracting affective information including symbols, as described in Embodiment 1 above, will be described.

[0066] Figure 16 is a block diagram (2) showing an example of the functional configuration of the information processing device 20. In Figure 16, the same reference numerals are used for functional parts that have the same functions as the information processing device 10 shown in Figure 1, while different reference numerals are used for functional parts that have different functions. Here, the explanation of functional parts that have the same functions as the information processing device 10 shown in Figure 1 will be omitted.

[0067] As shown in Figure 16, the information processing device 20 differs from the information processing device 10 shown in Figure 1 in that the control unit 25 further includes a sensory information extraction unit 25A.

[0068] The affective information extraction unit 25A has the function of extracting affective information, including symbols such as affective words. One aspect of this is that the affective information extraction unit 25A acquires information transmitted by participants in a meeting conducted by the conference system 3 (hereinafter referred to as "transmitted information") and situational information corresponding to said transmitted information. For example, transmitted information may include information that can be sensed when speaking, such as spoken text, as well as intonation, sound pressure, amount of silence, facial expressions, posture, gestures, body odor, physiological phenomena, and physical condition. Situational information may also include related information that, when considered together with transmitted information, may contribute to accurate understanding. For example, it may include participant profiles, the place and time of transmission, relationships between participants, the subject of communication (agenda and purpose if it is a meeting), communication history and event history leading up to the transmitted information, related external information (news, weather information, sensor data, etc.), physical information of participants (location, speed, posture, etc.), and biometric information of participants (brain waves, heart rate, etc.). Such situational information can be obtained, for example, from schedule information registered in the participant's scheduler, or by accepting user input for each item of situational information from the participant's user terminal. Then, the affective information extraction unit 25A inputs a prompt 51 containing the transmission information and situational information to the LLM 5, causing the LLM 5 to output the affective information.

[0069] Figure 17 shows an example of a prompt 51. As shown in Figure 17, the prompt 51 includes an instruction that tells the LLM 5 to perform the task of extracting affective information from transmitted information based on the contextual information. Furthermore, the prompt 51 is embedded with the results of the contextual information acquisition, as well as a log of the spoken text as an example of transmitted information to be used for extracting affective information. When such a prompt 51 is input to the LLM 5, affective information is output.

[0070] Here, we have given an example of having LLM5 extract emotional words, but the emotional information extraction unit 25A can also accept the designation of emotional words from the user terminals of participants in a meeting conducted by the conference system 3. Figure 18 is a diagram showing an example of a meeting screen 300. Figure 18 illustrates the meeting screen 300 displayed on the user terminal of participant B, one of the participants A and B in a meeting conducted by the conference system 3. As shown in Figure 18, the meeting screen 300 includes a participant display area 310 that displays information about the participants in the meeting conducted by the conference system 3, and a minutes display area 320 that displays the minutes of the meeting generated by a transcription function such as transcription. The system can also accept the designation of emotional words by accepting user input, such as range specification by drag and drop, from among the spoken text corresponding to the minutes of the meeting displayed in such a minutes display area 320. For example, in the example shown in Figure 18, the text "time machine" specified in the range indicated by the inverted display can be accepted as an emotional word. This allows recipients to designate words whose meaning they don't fully understand as "emotional words," and also allows the sender to designate words they feel are difficult for other participants to understand as "emotional words."

[0071] <Processing Flow> Figure 19 is a flowchart showing the procedure for extracting affective information. As shown in Figure 19, the affective information extraction unit 25A acquires transmitted information via the conference system 3 and the user terminals of participants in the conference conducted by the conference system 3, and also acquires status information corresponding to said transmitted information (step S201).

[0072] Next, the affective information extraction unit 25A inputs a prompt to the LLM 5 containing the transmission information and situation information acquired in step S201, causing the LLM 5 to extract symbols from the transmission information based on the situation information (step S202).

[0073] As a result, if a symbol is extracted (step S203: Yes), the affective information extraction unit 25A outputs the symbol extracted in step S202 to the element generation unit 15A (step S204), and the process ends. If no symbol is extracted (step S203: No), the process in step S204 is skipped, and the process ends.

[0074] <Summary of Embodiment 2> As described above, the information processing device 20 according to this embodiment inputs prompts embedded with transmission information and situation information to the LLM 5, causing the LLM 5 to extract symbols from the transmission information based on the situation information. Therefore, the information processing device 20 according to this embodiment has the effect of improving the accuracy of symbol extraction, including emotional words, in addition to the effects of the information processing device 10 according to Embodiment 1 described above.

[0075] <Application Example of Embodiment 2> In Embodiment 2 described above, an example was given in which situational information is used to extract affective information, but situational information can also be used to generate elements. Figure 20 is a diagram showing an example of a prompt 52. As shown in Figure 20, prompt 52 is superior to prompt 50 shown in Figure 5 in that more situational information is embedded in it. For example, an instruction sentence that instructs a task is given an instruction to consider situational information in the analysis of affective words X, and the result of acquiring situational information by the affective information extraction unit 25A is embedded. When such a prompt 52 is input to the LLM5, it is possible to output elements that reflect the situation of a dialogue held in a meeting, etc., as elements that serve as clues to the meaning indicated by affective words X.

[0076] <Embodiment 3> Next, an example of the functional configuration of the information processing device 30 according to Embodiment 3 will be described. In this embodiment, an example will be described in which symbols and elements included in the usage history related to the above-mentioned presentation function are presented if their similarity to the set of elements generated by the above-mentioned presentation function exceeds a threshold.

[0077] Figure 21 is a block diagram (3) showing an example of the functional configuration of the information processing device 30. In Figure 21, the same reference numerals are used for functional parts that have the same functions as the information processing device 20 shown in Figure 16, while different reference numerals are used for functional parts that have different functions. Here, the explanation of functional parts that have the same functions as the information processing device 20 shown in Figure 16 will be omitted.

[0078] As shown in Figure 21, the information processing device 30 differs from the information processing device 20 shown in Figure 16 in that the storage unit 33 stores a usage history DB (DataBase) 33A, and the control unit 35 further includes a usage history extraction unit 35A.

[0079] The usage history DB 33A is a database that stores a collection of usage history related to the above-mentioned presentation function. Figure 22 shows an example of the usage history DB 33A. As shown in Figure 22, the usage history DB 33A may be data to which items such as ID, symbol, axis, element, importance, subject of conversation, location, time, sender, receiver, and relationship between sender and receiver are associated. Of these, "ID" refers to an example of identification information for usage history. Also, "importance" refers to the importance specified by annotations for elements in which the meaning of a symbol is decomposed along an axis, such as a weight. For example, taking a usage history data entry identified by ID "1" as an example, the computer can be made to identify that the above-mentioned presentation function was used for the symbol "poor texture". Furthermore, the data entry allows the computer to identify that, among the three elements mapped onto the abstract axis, the element "The diagram is messy" is annotated with an importance level of "0," the element "The granularity of information is inconsistent" is annotated with an importance level of "30%," and the element "The flow of explanation is difficult to understand" is annotated with an importance level of "70%."

[0080] As merely an example, data entries in the usage history DB 33A may be generated when an element selection is accepted by the element presentation unit 15B. For example, the "ID" field can be assigned the number following the highest ID among the IDs already generated in the usage history DB 33A. Note that sequential numbering is merely an example, and the numbering method can be arbitrary as long as it allows for unique identification. In addition, the "Symbol" field may store emotional information extracted by the emotional information extraction unit 25A as a symbol. Furthermore, the "Axis" and "Element" fields may store elements generated for each axis by the element generation unit 15A. In addition, for "Importance," the element presentation unit 15B can assign a value corresponding to the weight annotated to the element's importance to the element whose selection is accepted, while assigning zero to the importance of elements whose selection is not accepted. Other data items such as "Topic of conversation," "Place," "Time," "Sender," "Receiver," and "Relationship between sender and receiver" may store situational information acquired by the emotional information extraction unit 25A.

[0081] The usage history DB 33A shown in Figure 22 is merely an example, and only some data items may be included in the data entries of the usage history DB 33A. For example, only elements generated from symbols may be stored in the usage history DB 33A. Furthermore, the usage history DB 33A may also include transmission information acquired by the affective information extraction unit 25A.

[0082] The usage history extraction unit 35A is a processing unit that extracts usage history entries stored in the usage history DB 33A, i.e., usage history entries that have a set of elements similar to the set of elements generated by the element generation unit 15A.

[0083] In one embodiment, the usage history extraction unit 35A acquires a set of elements generated by the element generation unit 15A. The usage history extraction unit 35A then calculates the similarity between the set of elements included in each usage history stored in the usage history DB 33A and the set of elements generated by the element generation unit 15A.

[0084] By way of example only, the similarity can be calculated by executing the following processing on each of the usage history stored in the usage history DB 33A and the generation result generated by the element generation unit 15A. For example, the usage history extraction unit 35A converts each element generated from one symbol into a vector value. Further, the usage history extraction unit 35A takes a set of elements generated from one symbol as a population, assigns annotation information of the element, for example, a weight corresponding to the importance, to the vector value of each element, and performs weighted addition, thereby calculating a representative vector representing the symbol. Then, the usage history extraction unit 35A calculates the cosine similarity between the representative vector of the usage history stored in the usage history DB 33A and the representative vector of the generation result generated by the element generation unit 15A, thereby calculating the similarity between symbols.

[0085] A specific example of the method for calculating the similarity between symbol A and symbol B is described below. For example, let the set of elements of symbol A be (a １ , a ２ , ..., a ｎ ), let the importance annotated to a ｉ be w(a ｉ ), and let the value obtained by converting a ｉ into a vector be α ｉ . Further, let the set of elements of symbol B be (b １ , b ２ , ..., b ｍ ), let the importance annotated to b ｊ be w(b ｊ ), and let the value obtained by converting b ｊ into a vector be β ｊ . In this case, the representative vector of symbol A can be calculated according to the following formula (1), and the representative vector of symbol B can be calculated according to the following formula (2). Further, the cosine similarity between the representative vector of symbol A and the representative vector of symbol B can be calculated according to the following formula (3).

[0086]

[0087] Then, the usage history extraction unit 35A outputs to the element presentation unit 15B any usage history stored in the usage history DB 33A whose similarity exceeds the threshold Th. When usage history is output from the usage history extraction unit 35A to the element presentation unit 15B, the element presentation unit 15B can present symbols, elements, etc., included in the usage history output by the usage history extraction unit 35A, along with the elements generated by the element generation unit 15A.

[0088] Here, as an example of similarity, we have given an example of calculating cosine similarity between representative vectors, but similarity calculations using dynamic time warping or co-occurrence matrix / latent semantic analysis (LSA) may also be used. Furthermore, the similarity may be calculated not only by comparing sets of elements or annotation information, but also by further comparing the transmission information, situation information, and symbols extracted by the affective information extraction unit 25A with the transmission information, situation information, and symbols included in the usage history.

[0089] <Processing Flow> Figure 23 is a flowchart showing the procedure for the usage history extraction process. As shown in Figure 23, the usage history extraction unit 35A obtains a set of elements generated from symbols by the element generation unit 15A, for example, the set of elements generated in step S102 shown in Figure 8 (step S301).

[0090] Next, the usage history extraction unit 35A calculates the similarity between the set of elements included in each usage history stored in the usage history DB 33A and the set of elements obtained in step S301 (step S302).

[0091] Then, the usage history extraction unit 35A outputs to the element presentation unit 15B (step S303) the usage history from the usage history DB 33A whose similarity calculated in step S302 exceeds the threshold Th, and terminates the process.

[0092] <Summary of Embodiment 3> As described above, the information processing device 30 according to this embodiment presents a set of symbols and elements included in the usage history included in the usage history DB 33A whose similarity to the set of elements generated by the element generation unit 15A exceeds a threshold. Therefore, according to the information processing device 30 according to this embodiment, in addition to the effects of the information processing device 20 according to Embodiment 2 described above, past real-world experiences are also presented as clues, thereby improving the accuracy of presenting the concept that the sender wants to express.

[0093] As an example of an application scenario to which the information processing device 30 according to the above embodiment 3 is applied, we will illustrate with application scenario 4.

[0094] (4) Application Scenario 4 As a fourth example, we will consider Application Scenario 4, which involves providing expressive support to improve the performance of collaborative work. Such Application Scenario 4 may have the following background. For example, in a dialogue between driver A and engineer B in motorsport, where they are considering machine settings to improve lap times, they want to evaluate how much of driver A's intuition is included in driver A's response to engineer B's questions. The information processing device 30 according to Embodiment 3 described above is applied under this background.

[0095] Figure 24 shows an example of the processing content for application scene 4. As shown in Figure 24, "Hmm, there is a rough feeling" is input as an example of text. This text is part of a conversation that took place in a scene where driver A communicates to engineer B his impressions of the machine's behavior during a test run on a circuit. With this input, the element generation unit 15A extracts the emotion word X "rough feeling" from the text "Hmm, there is a rough feeling" (step S1). Next, the element generation unit 15A decomposes the meaning of the emotion word X "rough feeling" along multiple axes (step S2). Then, the element presentation unit 15B receives an annotation for the element "abruptness" mapped to the abstract axis via the user terminal of driver A. After that, the element generation unit 15A receives an instruction to re-decompose the element "abruptness" on the spatial axis via the user terminal of engineer B. The element presentation unit 15B then receives an annotation for the element "moving to the left" mapped on the spatial axis by the re-decomposition via the user terminal of driver A (step S3). Subsequently, the usage history extraction unit 35A extracts usage history from the usage history DB 33A that has a set of elements similar to the set of elements obtained in the re-decomposition of step S3, and causes the element presentation unit 15B to display the symbols, etc., of the usage history (step S4). As a result, the selection result screen 210 shown in Figure 25 is displayed to the participant's user terminal.

[0096] Figure 25 shows an example of the selection result screen 210 in application scene 4. As shown in Figure 25, during the first decomposition, the selection result screen 211 is displayed, in which driver A has annotated "abruptness" among the elements arranged on the abstract axis in the order of "surprise," "dissatisfaction," "strong impact," "abruptness," "strong impact near the pelvis," and "physical experience of the hips collapsing." Subsequently, on the selection result screen 211, engineer B gives an instruction to re-decompose the element "abruptness" on the spatial axis, and the second decomposition is performed. That is, during the second decomposition, the selection result screen 212 is further displayed, in which driver A has annotated "progressing to the left" among the elements arranged on the spatial axis in the order of "upward behavior," "up and down behavior," "progressing to the left," and "progressing to the right." After that, a usage history having a set of elements similar to the set of elements included in the selection result screen 212 is extracted from the usage history DB 33A, and the usage history screen 213 related to that usage history is displayed. By sharing these selection result screens 211, 212, and usage history screen 213 with driver A and engineer B, the following effects can be obtained in application scenario 4. For example, it becomes possible to match the information with the detailed cues of driver A's past experiences, making it possible to share the subjective meaning that driver A wants to convey more accurately. This helps to improve the performance of collaborative work in application scenario 4, where expression support is provided to improve the performance of collaborative work.

[0097] <Embodiment 4> Next, an example of the functional configuration of the information processing device 40 according to Embodiment 4 will be described. In this embodiment, an example will be described in which the axis to be used for element generation by the element generation unit 15A is updated according to the evaluation result of the axis.

[0098] Figure 26 is a block diagram (4) showing an example of the functional configuration of the information processing device 40. In Figure 26, the same reference numerals are used for functional parts that have the same functions as the information processing device 30 shown in Figure 21, while different reference numerals are used for functional parts that have different functions. Here, the explanation of functional parts that have the same functions as the information processing device 30 shown in Figure 21 will be omitted.

[0099] As shown in Figure 26, the information processing device 40 differs from the information processing device 30 shown in Figure 21 in that the control unit 45 further includes an axis update unit 45A.

[0100] The axis update unit 45A has the function of updating the axes to be used for element generation by the element generation unit 15A. In one embodiment, the axis update unit 45A acquires the axes to be used for element generation by the element generation unit 15A. Subsequently, the axis update unit 45A evaluates each axis. As just one example, after the axis is presented by the element presentation unit 15B, the axis update unit 45A calculates an evaluation value for the spoken text related to each axis based on the magnitude of the index (TF-IDF) that evaluates the amount of information in the text. As another example, in a scene where multiple people work together, the axis update unit 45A calculates the evaluation value of the axis based on the performance of the task associated with the dialogue after the axis is presented by the element presentation unit 15B. As yet another example, after the axis is presented by the element presentation unit 15B, the axis update unit 45A can also receive subjective evaluations of each axis, such as questionnaires, from user terminals of participants, and calculate the evaluation value of each axis based on these subjective evaluations. As another example, the axis update unit 45A can also calculate evaluation values ​​for each axis based on the human's biological information after the axis is presented by the element presentation unit 15B. For example, evaluation values ​​for each axis can be calculated based on the ratio of alpha waves to beta waves in the electroencephalogram. As yet another example, the axis update unit 45A can also have a machine learning model evaluate the quality of the dialogue and behavior after the axis is presented by the element presentation unit 15B. After evaluation results are obtained for each axis in this way, the axis update unit 45A updates the axes to be used for element generation by the element generation unit 15A based on the evaluation results for each axis. For example, by inputting a prompt to the LLM 5 that contains the evaluation results for each axis and an instruction to filter the axes to be used for element generation using the evaluation results for each axis, the LLM 5 can be made to output the axes to be used for element generation by the element generation unit 15A. In addition, the axis update unit 45A can also narrow down the axes to those whose evaluation values ​​are above a threshold. Subsequently, the axis update unit 45A outputs to the element generation unit 15A a breakdown of the axes to be used for element generation by the element generation unit 15A, for example, the results of filtering and selection. Alternatively, axes specified via user terminals, such as those of meeting participants, may be output to the element generation unit 15A.

[0101] <Processing Flow> Figure 27 is a flowchart showing the procedure for the axis update process. As shown in Figure 27, the axis update unit 45A acquires axes to be used for element generation by the element generation unit 15A (step S401). Subsequently, the axis update unit 45A evaluates each axis acquired in step S401 (step S402).

[0102] Next, the axis update unit 45A updates the axes to be used by the element generation unit 15A for element generation based on the evaluation results of each axis (step S403). Finally, the axis update unit 45A outputs to the element generation unit 15A a breakdown of the axes to be used by the element generation unit 15A for element generation, for example, the results of filtering (step S404), and terminates the process.

[0103] <Summary of Embodiment 4> As described above, the information processing device 40 according to this embodiment updates the axis to be used for element generation by the element generation unit 15A according to the axis evaluation result. Therefore, the information processing device 40 according to this embodiment has the effect of generating elements corresponding to the optimized axis, in addition to the effects of the information processing device 30 according to Embodiment 3 described above, thereby further improving the accuracy of presenting the concept that the sender wants to express.

[0104] <Exhibition of Creative Ability> The matters described in Embodiments 1 to 4 above, such as specific examples of information to be transmitted, emotional language, axes, and elements, are merely examples and can be changed. In addition, the flowcharts described in Embodiments 1 to 4 above can also be modified in terms of the order of processing or by skipping some processing, as long as it is not inconsistent.

[0105] <System> Unless otherwise specified, the processing procedures, control procedures, specific names, and various data and parameters shown in the above documents and drawings may be changed at will. For example, one or more of the functional units of the information processing devices 10 to 40 may be composed of separate devices.

[0106] Furthermore, the components of each illustrated device are functionally conceptual and do not necessarily need to be physically configured as shown. In other words, the specific forms of distribution and integration of each device are not limited to those shown. That is, all or part of them can be functionally or physically distributed and integrated in any units according to various loads and usage conditions. Note that each configuration may also be a physical configuration.

[0107] Furthermore, the processing performed by the illustrated apparatus can be implemented, in whole or in part, by programs executed by hardware processors such as MPUs (Micro-Processing Units), CPUs (Central Processing Units), and GPUs (Graphics Processing Units), or by hardware using wired logic.

[0108] <Hardware> Next, we will describe an example of the hardware configuration of the information processing devices 10 to 40 described in Embodiments 1 to 4 above. For example, the information processing devices 10 to 40 can be implemented by installing a program that realizes the functions of the information processing devices 10 to 40 on a computer. For example, by having the computer run the above program, which is provided as packaged software or online software, the computer can be made to function as an information processing device 10 to 40. The computer referred to here includes desktop or notebook personal computers, rack-mounted server computers, etc. In addition, the computer category also includes smartphones, mobile phones and PHS (Personal Handyphone System) and other mobile communication terminals, as well as PDAs (Personal Digital Assistants). Furthermore, the functions of the information processing device 10 may be implemented on a cloud server.

[0109] An example of a computer that executes the above program (presented program) will be explained using Figure 28. As shown in Figure 28, the computer 1000 has, for example, a memory 1010, a CPU 1020, a hard disk drive interface 1030, a disk drive interface 1040, a serial port interface 1050, a video adapter 1060, and a network interface 1070. These components are connected by a bus 1080.

[0110] Memory 1010 includes ROM (Read Only Memory) 1011 and RAM (Random Access Memory) 1012. ROM 1011 stores, for example, a boot program such as BIOS (Basic Input Output System). The hard disk drive interface 1030 is connected to the hard disk drive 1090. The disk drive interface 1040 is connected to the disk drive 1100. The disk drive 1100 is used to insert a removable storage medium, such as a magnetic disk or an optical disk. The serial port interface 1050 is used to connect, for example, a mouse 1110 and a keyboard 1120. The video adapter 1060 is used to connect, for example, a display 1130.

[0111] Here, as shown in Figure 28, the hard disk drive 1090 stores, for example, the OS 1091, the application program 1092, the program module 1093, and the program data 1094. The storage unit 13 described in the above embodiment is equipped, for example, in the hard disk drive 1090 or the memory 1010.

[0112] Then, the CPU 1020 reads the program module 1093 and program data 1094 stored in the hard disk drive 1090 into the RAM 1012 as needed and executes the above-described procedures.

[0113] Furthermore, the program module 1093 and program data 1094 related to the above-mentioned program are not limited to being stored in the hard disk drive 1090, but may also be stored in a removable storage medium and read by the CPU 1020 via a disk drive 1100 or the like. Alternatively, the program module 1093 and program data 1094 related to the above-mentioned program may be stored in another computer connected via a network such as a LAN or WAN (Wide Area Network) and read by the CPU 1020 via a network interface 1070.

[0114] 3. Conference System 10, 20, 30, 40 Information Processing Device 11 Communication Control Unit 13, 33 Storage Unit 13A Model Information 33A Usage History DB 15, 25, 35, 45 Control Unit 15A Element Generation Unit 15B Element Presentation Unit 25A Affective Information Extraction Unit 35A Usage History Extraction Unit 45A Axis Update Unit

Claims

1. An information processing device characterized by comprising: a generation unit that generates elements that serve as clues to the meaning of words used in linguistic communication, along axes implicitly used by humans when verbalizing complex concepts; and a presentation unit that associates and presents elements corresponding to each axis.

2. The information processing apparatus according to claim 1, characterized in that the generation unit inputs prompts into a large-scale language model in which emotional words contained in spoken text are embedded, thereby causing the large-scale language model to generate elements that serve as clues to the meaning of the emotional words.

3. The information processing apparatus according to claim 1, characterized in that the generation unit generates the elements using at least one axis from among an axis defined by abstraction level, an axis defined by human physical sensation, an axis defined by spatial positional relationship, an axis defined by time series, and an axis defined by purpose or function.

4. The information processing apparatus according to claim 1, characterized in that the generation unit further generates a second element which serves as a clue to the meaning of the first element from the first element generated by the generation unit.

5. The information processing apparatus according to claim 1, characterized in that the display unit accepts a selection from among the elements presented for each axis, and further presents the elements for which the selection has been accepted.

6. The information processing device according to claim 1, further comprising an extraction unit that extracts a history from which a set of elements similar to the set of elements generated by the generation unit is associated with the words used in the language communication, wherein the presentation unit presents information about the history extracted by the extraction unit together with the set of elements generated by the generation unit.

7. A presentation method performed by an information processing device, comprising: a generation step of generating elements that serve as clues to the meaning of words used in linguistic communication, along axes implicitly used by humans when verbalizing complex concepts; and a presentation step of associating and presenting elements corresponding to each axis.

8. A presentation program for causing a computer to perform a generation step of generating elements that serve as clues to the meaning of words used in linguistic communication, along axes implicitly used by humans when verbalizing complex concepts, and a presentation step of associating and presenting elements corresponding to each axis.

Images