Sensitivity conversion method, sensitivity conversion device, and sensitivity conversion program

Abstract

A storage unit (14) stores history data (14a) in which information indicating a past situation of a speaker and an interlocutor is associated with an expression expressing sensitivity in that situation. An extraction unit (15b) extracts an expression expressing sensitivity from a phrase uttered by the speaker. An estimation unit (15c) uses the history data (14a) to estimate information expressing a situation corresponding to the extracted expression expressing sensitivity. An identification unit (15d) uses the history data (14a) to identify an expression expressing the sensitivity of the interlocutor corresponding to the estimated information expressing a situation.

Description

[Technical Field]

[0001] The present invention relates to a sensitivity conversion method, a sensitivity conversion device, and a sensitivity conversion program. [Background technology]

[0002] Conventionally, there are known techniques for converting language expressions to suit the other party in a dialogue. For example, there is known a technique for converting text from one language to another language while preserving its meaning, such as in multilingual translation (see Patent Document 1). There is also known a technique for converting a certain text into an expression that is more relatable (see Non-Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent Publication No. 2021-039501 [Non-patent literature]

[0004] [Non-Patent Document 1] A. Sharma et al, “Towards Facilitating Empathic Conversations in Online Mental Health Support: A Reinforcement Learning Approach”, 2021 Summary of the Invention [Problem to be solved by the invention]

[0005] However, with conventional technology, it is difficult to resolve the gap in understanding and recognition between the speaker and the receiver during dialogue-based communication. For example, there may be a difference in the degree to which a speaker says "it feels good" and the receiver feels "it feels good." This gap in understanding and recognition is thought to be caused by differences in the sensibilities of the speaker and the receiver. Sensitivity differs from person to person and is formed by various factors, such as a person's innate personality, past experiences, and relationship with the other person. Gaps in understanding and recognition are a cause of miscommunication and can be a major problem in conversations such as business meetings and casual conversations.

[0006] The present invention has been made in view of the above, and has as its object to eliminate the gap in understanding and recognition between a speaker and a conversational partner in conversational communication. [Means for solving the problem]

[0007] In order to solve the above-mentioned problems and achieve the object, the sensitivity conversion method of the present invention is a sensitivity conversion method executed by a sensitivity conversion device, wherein the sensitivity conversion device has a memory unit that stores history data that associates information representing past situations of a speaker and an interlocutor with expressions that represent the sensitivity in the situations, and is characterized by including an extraction step of extracting expressions that represent the sensitivity from sentences spoken by the speaker, an estimation step of using the history data to estimate information representing the situation that corresponds to the extracted expressions that represent the sensitivity, and an identification step of using the history data to identify expressions that represent the sensitivity of the interlocutor that correspond to the estimated information representing the situation. [Effects of the Invention]

[0008] According to the present invention, it is possible to eliminate the gap in understanding and recognition between a speaker and a person speaking in dialogue. [Brief explanation of the drawings]

[0009] [Figure 1] FIG. 1 is a schematic diagram illustrating the general configuration of a sensitivity conversion device. [Figure 2]FIG. 2 is a diagram illustrating an example of the data structure of attribute data. [Figure 3] FIG. 3 is a diagram for explaining the processing of the sensitivity conversion device. [Figure 4] FIG. 4 is a flowchart showing the procedure of the emotion conversion process. [Figure 5] FIG. 5 is a diagram illustrating a computer that executes the emotion conversion program. DETAILED DESCRIPTION OF THE INVENTION

[0010] Hereinafter, an embodiment of the present invention will be described in detail with reference to the drawings. Note that the present invention is not limited to this embodiment. In addition, in the description of the drawings, the same parts are designated by the same reference numerals.

[0011] [Overview of the emotion conversion device] In verbal communication, when a speaker inputs text that verbalizes his or her sensibilities in a certain situation, the sensibility conversion device of this embodiment converts the text into text that will give the recipient the same or similar sensibility as the speaker and outputs it.

[0012] The emotion conversion device is used, for example, when expressing the emotion that a speaker feels about a certain situation itself. The emotion conversion device may also be used when a situation changes due to some trigger or over time, predicting a future situation, and expressing the emotion felt about the predicted future situation.

[0013] For example, when playing a game with multiple people as a team, it can be used to express one's feelings about the current situation of the game, or when predicting the possible future actions of one's own team or rival team and expressing one's feelings about the predicted future situation. Or when watching a sports game with multiple people, it can be used to express one's feelings about the current situation, or when imagining the next attack that the supporters or opponents may take and expressing one's feelings about the possible future situation.

[0014] Here, expressions expressing emotions (hereinafter referred to as emotion expressions) are verbal expressions of the emotions felt by the speaker based on input situations or triggers. For example, expressions such as "Oh no, we're going to lose" expressed after observing the game situation and the opponent's attack are examples of such expressions. Even if people have the same emotions, the resulting verbalized emotion expressions are likely to differ depending on the person. For example, a timid person will frequently use the phrase "we're going to lose" when they are even slightly behind, while a confident person will say "it'll be fine" when they are only slightly behind, and will not say "we're going to lose" until they are one step away from losing. In this case, if one interacts with another person without knowing their personality, even if the timid person says "we're going to lose," the confident person will not understand why they feel so behind, which could lead to miscommunication.

[0015] To reduce such miscommunication, the emotion conversion device can instruct a confident person, for example, that "When the other person just said, 'I'm going to lose,' to you, it means 'It'll be fine.'" This allows people to understand the differences in how they perceive a situation, the range of linguistic expression, and the other person's personality, leading to greater understanding of others and diversity and reducing miscommunication.

[0016] The emotion conversion device described below is applied to a situation where the content of the conversation is Shogi, and two players, A and B, form a team and converse with each other to consider their next move. Note that the conversation within the team cannot be heard by the opponent, for example, in a remote Shogi game using a computer or when the opponent is an AI.

[0017] [Configuration of the emotion conversion device] Fig. 1 is a schematic diagram illustrating the general configuration of a sensitivity conversion device. As illustrated in Fig. 1, a sensitivity conversion device 10 of this embodiment is realized by a general-purpose computer such as a personal computer, and includes an input unit 11, an output unit 12, a communication control unit 13, a storage unit 14, and a control unit 15.

[0018] The input unit 11 is realized using input devices such as a keyboard and a mouse, and inputs various instruction information such as starting processing to the control unit 15 in response to input operations by the user. The output unit 12 is realized by a display device such as a liquid crystal display, a printing device such as a printer, an information communication device, etc. The communication control unit 13 is realized by a NIC (Network Interface Card) or the like, and controls communication via a network between the control unit 15 and external devices such as a server or a management device that manages various data.

[0019] The storage unit 14 is realized by a semiconductor memory element such as a RAM (Random Access Memory) or a flash memory, or a storage device such as a hard disk or an optical disk. The storage unit 14 may be configured to communicate with the control unit 15 via the communication control unit 13. In this embodiment, the storage unit 14 stores, for example, history data 14a used in the affective conversion process described below, a first model 14b, a second model 14c generated and updated in the affective conversion process, and the like.

[0020] Here, Fig. 2 is a diagram illustrating the data structure of history data. As shown in Fig. 2, history data 14a associates information representing past situations of the speaker and interlocutor with expressions expressing the emotions in those situations. In Fig. 2, information such as the state of the board (winning rate), the number of moves made, and the number of pieces in hand is recorded as past situations in shogi. In addition, emotional expressions expressing the emotions of players a and b, such as "scary" and "dangerous," corresponding to these situations are associated.

[0021] In addition, in the history data 14a, the information representing the situation may be, for example, information that changes over time. Alternatively, the information representing the situation may be information that changes depending on a trigger that changes the situation. Furthermore, the information representing the situation may be an evaluation value obtained by AI.

[0022] For example, in games or sports, the situation can be the score or loss, and the trigger that changes the situation can be the opponent's attacking pattern. The situation and the trigger that changes the situation represent the environment that caused the speaker to feel something in their brain, and it is thought that the speaker develops emotions based on this and then verbalizes them in speech. The speaker's speech can be pre-entered as text, as in text chat, or can be speech that has been recognized or manually transcribed.

[0023] 3 is a diagram for explaining the processing of the sentiment conversion device. In the current situation x of a shogi game, the sentiment conversion device 10 of this embodiment converts the sentiment felt by player a immediately after the opponent plays the next move y, such as "3-6-Fu," as a trigger to change the situation x, into the sentiment of player b and conveys it to player b. In the example shown in FIG. 3, the sentiment expression of player a, "scary," is converted into the sentiment expression of player b, "very dangerous."

[0024] The current situation x in shogi represents the objectively observable current situation, and is expressed as a vector combining multiple values, such as the win rate value evaluated by the AI ​​for the current score, the position of the pieces, the type and number of pieces in hand, and the game record so far.

[0025] Returning to the explanation of FIG. 1, the control unit 15 is realized using a CPU (Central Processing Unit), an NP (Network Processor), an FPGA (Field Programmable Gate Array), or the like, and executes a processing program stored in memory. As a result, the control unit 15 functions as an acquisition unit 15a, an extraction unit 15b, an estimation unit 15c, an identification unit 15d, and a conversion unit 15e, as exemplified in FIG. 1. Note that these functional units may each be implemented in different hardware. For example, the acquisition unit 15a may be implemented in hardware different from the other functional units. The control unit 15 may also include other functional units.

[0026] The acquisition unit 15a acquires an utterance sentence that verbalizes the sensibility of a speaker. For example, the acquisition unit 15a acquires an utterance sentence t that verbalizes the sensibility felt by the player a. a Get.

[0027] The extraction unit 15b extracts expressions expressing sensitivities (emotional expressions) from sentences uttered by a speaker. Specifically, the extraction unit 15b divides the utterance sentence into words by morphological analysis and extracts emotion expressions from among the words. The extraction unit 15b may, for example, refer to a dictionary of words that are to be emotion expressions constructed in advance and extract words that fall under the dictionary as emotion expressions. Alternatively, the extraction unit 15b may extract adjectives, which often express the degree of something or a way of feeling, as emotion expressions.

[0028] Specifically, the extraction unit 15b extracts the input utterance sentence t a From the above, the sensibility expression w of the player a's hand y in the situation x a,x,y Output.

[0029] The extraction unit 15b may transfer all words divided by the morphological analysis to the estimation unit 15c described below without extracting affective expressions. In this case, the estimation unit 15c may process only affective expressions.

[0030] The estimation unit 15c estimates information representing a situation corresponding to the extracted affective expression using the history data 14a.

[0031] For example, the estimation unit 15c estimates information representing a situation corresponding to the extracted affective expression using the first model 14b trained using the historical data 14a so as to output information representing the corresponding situation when the speaker's affective expression is input.

[0032] That is, the estimation unit 15c uses the first model 14b trained using the historical data 14a indicating what kind of utterances the speaker made in the past and in what kind of situations, to estimate what kind of situation the speaker is describing with respect to the input emotional expression word or group of words.

[0033] The identification unit 15d uses the history data 14a to identify the emotional expression of the interlocutor corresponding to the information representing the estimated situation.

[0034] For example, when information representing a situation is input, the identification unit 15d identifies the emotional expression of the interlocutor corresponding to the information representing the estimated situation using the second model 14c learned using the historical data 14a, so as to output the emotional expression of the interlocutor in the situation.

[0035] In other words, the identification unit 15d uses a second model 14c trained using historical data 14a indicating what kind of utterances the recipient (interlocutor) made in the past under what circumstances, to predict, in words or groups of words, how the recipient's interlocutor would verbalize the estimated situation.

[0036] In the sensitivity conversion device 10 of this embodiment, the history data 14a contains a set of a game situation representing the game board and the player's sensitivity expression for the next move for past games between players a and b. In the example shown in FIG. 2, the value representing the situation is an evaluation value such as a win rate evaluated by an AI for the current game board and expressed as a numerical value of 0-1. In this embodiment, an AI is used as the opponent, but the AI ​​that calculates the evaluation value and the opponent's AI may be the same or different.

[0037] The evaluation value p(y|x) of the next move y in a given situation x will be the same for all players in the same situation. However, in reality, there will be moves that players like, dislike, or will not consider as next candidates depending on their personality, values, preferred playing style, shogi school, cultural background, etc. Therefore, an evaluation value P(y|x, a) weighted for each player may be used in addition to the evaluation value calculated by AI. For example, for player a who prefers an offensive strategy, a certain number may be added to p(y|x) when the board is easy to attack, and p(y|x) may be multiplied by a decimal point less than 1 when the board is difficult to attack.

[0038] For each state of a game played by a player a in the past, the evaluation value can be calculated by taking into account the evaluation value annotated by the player himself or herself and the final game results. shogi,a (x, y) can be defined and used to calculate the evaluation value P(y|x, a). Alternatively, instead of calculating for each player, we can use a function M for all players with game data. shogi,all You can also calculate the evaluation value using (x, y). However, you can also calculate the evaluation value using M shogi,player By calculating (x, y) as an evaluation value, it becomes possible to express the subtle differences in sensibility of each player in different linguistic expressions.

[0039] In addition, the spaces of players a and b are formed from the experience of each player's previous games, so the data is sparse, and the possibility of the same situation x occurring again is low, so it may be impossible to calculate the evaluation value. Therefore, for example, personas can be defined by user attributes such as school, rank, years of experience, and age, and the evaluation value M can be calculated for players grouped by persona. shogi,a~ (x, y) may be defined. This makes it possible to calculate the evaluation value using history data 14a of past games of people with similar sensibilities to player a. Furthermore, if there is no matching situation x in the past, a value with little influence, such as the median of the possible values, may be used as the evaluation value.

[0040] Then, in the emotion conversion process, the emotion expression of player a for hand y in situation x is converted into text that allows player b to have the same or similar emotion as player a.

[0041] Here, the emotional expression by player a is expressed by the following equation (1) using a mapping function F that differs for each player.

[0042]

number

[0043] Function F a is a linguistic expression f a That is, F a is equivalent to defining as a function the process by which player A inputs information about the current situation into his brain and verbalizes what he feels.

[0044] Furthermore, the emotional expression by player b is expressed by the following equation (2).

[0045]

number

[0046] Function F b is the winning probability P' of player b for move y in situation x, and the linguistic expression f b That is, F b is equivalent to defining the process of verbalizing what player B feels in his mind as a function.

[0047] where F a Inverse function F a -1 can be considered as a function that returns the verbalized emotion expression to the original situation that triggered the player a to have that emotion. a -1 To input the sensibility expression of player a, a -1 (w a,x,y ) This can be said to represent the situation that is the source of player A's emotional expression.

[0048] This situation is expressed as a function F b Entering in F b (F a -1 (w a,x,y)) This can be said to be an emotional expression that player b verbalizes when player a inputs the situation that player a perceives into player b's mind. Therefore, the following formula (3) holds for the text converted into player b's emotional expression.

[0049]

number

[0050] In this way, the emotional expression uttered by player a about situation x and move y can be input and converted into an emotional expression that allows player b to have the same emotion.

[0051] where F a -1 , F b is modeled by machine learning using historical data 14a. a -1 is a first model 14b that is trained to input the emotional expression of the player a and output an evaluation value P(y|x, a) that represents the situation. b is the evaluation value P'(y|x, b) of a situation that player b experienced in the past, and the emotional expression w b,x,y and a second model 14c trained to output

[0052] The machine learning method is not particularly limited. a -1 The sentiment expressions entered in may be just the Bag-of-Words expressions, or may include additional variables that explain the meaning and usage of the words used as sentiment expressions, such as numbers or vectors that express the meaning of the words, or numbers that indicate positive / negative.

[0053] The conversion unit 15e converts the utterance sentence using an expression that expresses the sensibility of the identified interlocutor. For example, the conversion unit 15e rewrites only the part of the speaker's utterance that has been converted as an emotional expression, and formats it so that it is grammatically valid. In other words, the conversion unit 15e formats the emotional expression output from the identification unit 15d so as not to impair the original context, and presents it to player B.

[0054] At this time, the conversion unit 15e may, for example, replace a part of the original utterance sentence with the converted affective expression so as to have natural grammar and present it. Alternatively, the conversion unit 15e may display the converted affective expression in parallel with the affective expression part of the original utterance sentence. Furthermore, the conversion unit 15e may present the converted affective expression in real time, or may perform affective conversion processing in advance and present it as feedback when reviewing the utterance later.

[0055] [Sensation conversion processing] Next, a description will be given of the emotion conversion process performed by the emotion conversion device 10. Fig. 4 is a flowchart showing the procedure of the emotion conversion process. The flowchart in Fig. 4 starts, for example, when an input is made to instruct the start of the process.

[0056] First, the acquisition unit 15a acquires an utterance sentence that verbalizes the speaker's sensibilities. Then, the extraction unit 15b extracts expressions that express sensibilities (sentimental expressions) from the utterance sentence (step S1). Specifically, the extraction unit 15b divides the utterance sentence into words by morphological analysis, and extracts sentimental expressions from the words.

[0057] Next, the estimation unit 15c estimates information representing a situation corresponding to the extracted affective expression using the history data 14a (step S2). For example, the estimation unit 15c estimates information representing a situation corresponding to the extracted affective expression using the first model 14b trained using the history data 14a so as to output information representing the corresponding situation when an affective expression of a speaker is input.

[0058] That is, the estimation unit 15c uses the first model 14b trained using the historical data 14a indicating what kind of utterances the speaker made in the past and in what kind of situations, to estimate what kind of situation the speaker is describing with respect to the input emotional expression word or group of words.

[0059] Furthermore, the identification unit 15d identifies an affective expression of the interlocutor corresponding to the information representing the estimated situation, using the history data 14a (step S3). For example, when information representing a situation is input, the identification unit 15d identifies an affective expression of the interlocutor corresponding to the information representing the estimated situation, using the second model 14c trained using the history data 14a, so as to output an affective expression of the interlocutor in the situation.

[0060] In other words, the identification unit 15d uses a second model 14c trained using historical data 14a indicating what kind of utterances the recipient (interlocutor) made in the past under what circumstances, to predict, in words or groups of words, how the recipient's interlocutor would verbalize the estimated situation.

[0061] Then, the conversion unit 15e converts the utterance sentence using an expression that expresses the specified interlocutor's sensibility (step S4). For example, only the part of the speaker's utterance sentence that has been converted as an emotional expression is rewritten and formatted so that the grammar is valid. That is, the conversion unit 15e formats the emotional expression output from the specification unit 15d so as not to impair the original context, and presents it to the interlocutor. This completes a series of emotion conversion processes.

[0062] [effect] As described above, in the sensitivity conversion device 10 of this embodiment, the storage unit 14 stores history data that associates information representing past situations of a speaker and interlocutors with expressions representing the sensibilities in those situations. The extraction unit 15b extracts expressions representing sensibilities from sentences uttered by the speaker. The estimation unit 15c estimates information representing a situation corresponding to the extracted expressions representing the sensibilities, using the history data 14a. The identification unit 15d identifies expressions representing the interlocutors' sensibilities that correspond to the estimated information representing the situation, using the history data 14a.

[0063] Specifically, the estimation unit 15c estimates information representing a situation corresponding to the extracted expression representing the sensibility using a first model 14b trained using the history data 14a so as to output information representing the corresponding situation when an expression representing the speaker's sensibility is input. Furthermore, the identification unit 15d identifies an expression representing the interlocutor's sensibility corresponding to the estimated information representing the situation using a second model 14c trained using the history data 14a so as to output an expression representing the interlocutor's sensibility in the situation when information representing the situation is input.

[0064] In this way, the sensitivity conversion device 10 converts the sensitivity expressed by the speaker into an expression that the interlocutor can feel in the same way, using data on the experiences of the speaker and interlocutor in the dialogue, the experiences of personas with similar sensitivities, and information that represents the current situation, such as evaluation values ​​using AI in competitive games like shogi.

[0065] This allows the interlocutor to more accurately grasp the speaker's feelings about the same situation, making it possible to communicate without any gaps in understanding or recognition. For example, when devising a strategy to win a game, by aligning each other's perception of the game situation, it becomes possible to align perspectives and communicate efficiently. Furthermore, by accurately grasping the linguistic expressions used by players and having the person in charge of planning the team's strategy faithfully grasp the players' sensibilities and reflect them in the strategy, the winning rate in the game increases. In this way, the sensibility conversion device 10 makes it possible to eliminate gaps in understanding or recognition between the speaker and the interlocutor in dialogue-based communication.

[0066] Furthermore, the information representing the situation may be information that changes over time. Alternatively, the information representing the situation may be information that changes depending on a trigger that changes the situation. This allows the emotion conversion device 10 to convert the emotion expression even when the situation changes.

[0067] Furthermore, evaluation values ​​by AI may be used as information representing the situation. In this way, by using objective evaluation values ​​by AI, etc. as information representing the situation, the situation can be expressed succinctly. Furthermore, if an evaluation value by AI can be obtained in a situation, it becomes possible to convert sensibilities even in different scenes. For example, if AI evaluation values ​​can be obtained for both shogi and sports, it becomes possible to convert words spoken in shogi into words spoken during a sports match.

[0068] Furthermore, the conversion unit 15e converts the utterance sentence using an expression that expresses the sensibility of the identified interlocutor, thereby enabling the interlocutor to accurately grasp the sensibility expression of the speaker and eliminate any discrepancy in understanding or recognition between the speaker and the interlocutor.

[0069] [program] A program describing the processing executed by the emotion conversion device 10 according to the above embodiment in a computer-executable language can also be created. In one embodiment, the emotion conversion device 10 can be implemented by installing an emotion conversion program that executes the above emotion conversion processing as package software or online software on a desired computer. For example, by causing an information processing device to execute the emotion conversion program, the information processing device can function as the emotion conversion device 10. Other examples of information processing devices include mobile communication terminals such as smartphones, mobile phones, and PHSs (Personal Handyphone Systems), as well as slate terminals such as PDAs (Personal Digital Assistants). The functions of the emotion conversion device 10 may also be implemented on a cloud server.

[0070] 5 is a diagram showing an example of a computer that executes a sensitivity conversion program. The computer 1000 includes, 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.

[0071] The memory 1010 includes a ROM (Read Only Memory) 1011 and a RAM 1012. The ROM 1011 stores, for example, a boot program such as a BIOS (Basic Input Output System). The hard disk drive interface 1030 is connected to a hard disk drive 1031. The disk drive interface 1040 is connected to a disk drive 1041. A removable storage medium such as a magnetic disk or optical disk is inserted into the disk drive 1041. The serial port interface 1050 is connected to, for example, a mouse 1051 and a keyboard 1052. The video adapter 1060 is connected to, for example, a display 1061.

[0072] Here, the hard disk drive 1031 stores, for example, an OS 1091, an application program 1092, a program module 1093, and program data 1094. Each piece of information described in the above embodiment is stored in the hard disk drive 1031 or memory 1010, for example.

[0073] The emotion conversion program is stored in the hard disk drive 1031 as a program module 1093 in which instructions to be executed by the computer 1000 are written. Specifically, the program module 1093 in which each process executed by the emotion conversion device 10 described in the above embodiment is written is stored in the hard disk drive 1031.

[0074] Furthermore, data used in information processing by the emotion conversion program is stored as program data 1094, for example, in the hard disk drive 1031. Then, the CPU 1020 reads the program module 1093 and program data 1094 stored in the hard disk drive 1031 into the RAM 1012 as necessary, and executes each of the above-described procedures.

[0075] The program module 1093 and program data 1094 related to the emotion conversion program are not limited to being stored in the hard disk drive 1031, but may be stored in a removable storage medium and read by the CPU 1020 via the disk drive 1041, etc. Alternatively, the program module 1093 and program data 1094 related to the emotion conversion program may be stored in another computer connected via a network such as a LAN (Local Area Network) or a WAN (Wide Area Network), and read by the CPU 1020 via the network interface 1070.

[0076] Although the present invention has been described above as an embodiment, the present invention is not limited to the description and drawings that form part of the disclosure of the present invention. In other words, other embodiments, examples, and operational techniques that can be made by those skilled in the art based on the present invention are all included in the scope of the present invention.

[0077] For example, in the above embodiment, a game example was used to clarify the issue, but it is possible to model communication with a history that generally involves consensus building using a similar framework. Instead of AI, it is possible to use the degree of agreement with the company's decision-making policy, the selection tendencies of ordinary people, etc. [Explanation of symbols]

[0078] 10 Sensibility Conversion Device 11 Input section 12 Output section 13 Communication control section 14 Storage section 14a Historical Data 14b First Model 14c Second Model 15 Control Unit 15a Acquisition part 15b Extraction part 15c Estimation part 15d Specific section 15e conversion unit

Claims

1. A sentiment conversion method executed by a sentiment conversion device, the emotion conversion device has a storage unit that stores history data in which information representing past situations of a speaker and a conversation partner is associated with expressions representing emotions in the situations; an extraction step of extracting expressions expressing sensibilities from sentences uttered by a speaker; an estimation step of estimating information representing a situation corresponding to the extracted expression representing the sensibility using the history data; a step of identifying an expression that expresses the sensibility of the interlocutor corresponding to the information that expresses the estimated situation using the history data; A method for converting emotions, comprising:

2. the estimation step estimates information representing a situation corresponding to the extracted expression expressing the sensibility by using a first model trained using the history data so as to output information representing a corresponding situation when an expression expressing the sensibility of a speaker is input; the identifying step identifies an expression expressing the sensibility of the interlocutor corresponding to the estimated information expressing the situation using a second model trained using the history data, so as to output an expression expressing the sensibility of the interlocutor in the situation, when information expressing a situation is input; The method for converting perceptions according to claim 1 .

3. 2. The method according to claim 1, wherein the information representing the situation is information that changes in a time series.

4. 2. The method for converting perceptions according to claim 1, wherein the information representing the situation is information that changes in response to a trigger that changes the situation.

5. The method for converting perceptions according to claim 1, wherein the information representing the situation is an evaluation value obtained by AI.

6. The sensitivity conversion method according to claim 1 , further comprising a conversion step of converting the utterance sentence using an expression that expresses the specified sensitivity of the interlocutor.

7. a storage unit that stores history data in which information representing past situations of a speaker and a conversation partner is associated with expressions representing feelings in the situations; an extraction unit that extracts expressions expressing sensitivities from sentences uttered by a speaker; an estimation unit that estimates information representing a situation corresponding to the extracted expression representing the sensibility using the history data; an identification unit that identifies an expression that expresses the sensibility of the interlocutor corresponding to the information that expresses the estimated situation, using the history data; A sensitivity conversion device comprising:

8. A sensitivity conversion program for causing a computer to execute the sensitivity conversion method according to any one of claims 1 to 6.

Images