Game system

Abstract

To provide a technique allowing one to enjoy development of a game by an atypical conversation in a system for operating a player character (PC) in space in which a non-player character (NPC) appears.SOLUTION: A game system includes: a learned model storage section (113) for storing a learned model that mechanically learns feeling data expressing conversation data and feelings following the conversation by a value of a feeling parameter, and generates a replay script and a value of a feeling parameter; an action information storage section (114) for storing action information with which a value of a feeling parameter and action of NPC are correlated; a reaction generation section (23) for generating a replay script and a value of a feeling parameter by inputting an input script to a learned model; script generation sections (33, 44) for outputting a replay script; and an action determination section (24) for determining action of an NPC for operation by comparing a generated value of a feeling parameter with action information.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a game system used for playing a game in a game space constituted by a virtual space or the like. [Background technology]

[0002] Conventionally, there have been known games in which a plurality of characters appearing in a virtual space converse with each other while the game progresses. For example, Patent Document 1 describes a game device in which a player character operated by a player can play a game in which the player character converses with a plurality of companion characters operated by a computer while adventuring in a virtual space.

[0003] In this game, for each of a plurality of companion characters, a conversation table is prepared in advance, which contains the companion character's lines (starting lines) that serve as the starting point for activating the companion character's special move, and the companion characters' lines, including lines from other companion characters that affirm or deny the activation of the special move (these are called "supporting lines"). When one of the companion characters utters the starting line in a battle scene in the game, the supporting lines of the other companion characters are randomly selected from those recorded in the conversation table, and if the selected supporting line affirms the activation of the special move, the special move of the companion character that uttered the starting line is activated. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2019-202079 A [Non-patent literature]

[0005] [Non-Patent Document 1] Hirofumi Hori, "AI appears to have emotions", [online], February 21, 2023, [Retrieved September 20, 2023], Internet<URL:https: / / note.com / horiega / n / nfc71df48cb7b> [Non-Patent Document 2] "What is a Large-Scale Language Model (LLM)? A thorough explanation of the types, applications, and issues", [online], August 31, 2023, Ice Smiley Co., Ltd., [Searched September 20, 2023], Internet<URL:https: / / aismiley.co.jp / ai_news / what-is-large-language-models / > [Non-Patent Document 3] "VOICEVOX", [online], June 2, 2023, Pixiv Encyclopedia, [Searched September 20, 2023], Internet<URL:https: / / dic.pixiv.net / a / VOICEVOX#h2_5> Summary of the Invention [Problem to be solved by the invention]

[0006] In the game device described in Patent Document 1, whether or not a friend character's special move is activated is determined by the randomly selected auxiliary lines of other friend characters, so that players can enjoy a non-standard game development. However, since the opening lines and auxiliary lines uttered by the friend characters are prepared in advance in a conversation table, there is a problem that players get bored of these lines uttered by the friend characters as they play the game repeatedly. Patent Document 1 also describes that the content of conversations for player characters and enemy characters may also be determined by the conversation table, but does not provide any specific examples.

[0007] The problem that the present invention aims to solve is to provide a technology in a game system in which a player controls a player character to play a game in a game space in which non-player characters controlled by a computer appear, whereby the player can enjoy the game development through non-standard conversations between the player character and the non-player character. [Means for solving the problem]

[0008] The present invention, which has been made to solve the above problems, is a system in which a player controls a player character to play a game in a game space in which non-player characters controlled by a computer appear, comprising: a display data storage unit in which display data of the game space, display data of the player character, and display data of the non-player character are stored; a trained model created by machine learning learning data including conversation data of a conversation consisting of an input line and a reply line, and emotion data expressing emotions arising from the conversation by emotion parameter values, the trained model storing a trained model that generates a reply line and the emotion parameter values ​​in response to an input of an input line; a behavior information storage unit in which behavior information that associates the emotion parameter value with the behavior of the non-player character is stored; a dialogue input unit for inputting dialogue of the player character; a reaction generation unit that inputs the lines of the player character input to the line input unit as input lines to the trained model to generate a response line and a value of the emotion parameter; a dialogue output unit that outputs the response dialogue generated by the trained model by the reaction generation unit as a dialogue of the non-player character; a behavior determination unit that determines a behavior of the non-player character in accordance with the value of the emotion parameter generated in the trained model by the reaction generation unit in light of the behavior information stored in the behavior information storage unit, and causes the non-player character to act in the game space; The present invention is characterized by comprising:

[0009] The trained model is, for example, a large language model (LLM) that has already been used in chatbots, etc. As described in Non-Patent Documents 1 and 2, such a large language model can generate a response line (response line) to an input line (input line) to establish a natural conversation between people, and generate values ​​of emotion parameters associated with such a conversation.

[0010] In the game system according to the present invention, a player inputs lines of a player character that the player operates into a line input unit. This input may be text input, voice input, or both. When the lines of the player character are input, the reaction generation unit inputs the input lines into a trained model to generate a reply line and an emotion parameter value. The generated lines are output by the line output unit as lines of a non-player character. This output may also be text output, voice output, or both. In addition, the action determination unit determines the action of the non-player character based on the generated emotion parameter value in light of previously prepared action information, and causes the non-player character to move in the virtual space.

[0011] In the game system according to the present invention, a response line is generated in response to any line spoken by the player character, so that a non-standard conversation can be enjoyed between the player character and the non-player character. Also, an emotion parameter value is generated in response to the line spoken by the player character, and the action of the non-player character is determined in response to the value, so that a different game development can be enjoyed in response to the line spoken by the player character.

[0012] The game system according to the present invention can be applied to various kinds of games. For example, when configured as a game system for playing a battle type game, the value of the emotion parameter can be associated with the level of ability of a non-player character, and the difficulty level when a player character battles a non-player character can be changed.

[0013] In the game system according to the present invention, a plurality of the emotion parameters may be used. By using a plurality of emotion parameters, the behavior of the non-player characters can be set in a more complex manner, and a more diverse game development can be enjoyed. Effect of the Invention

[0014] By using the game system of the present invention, when a player controls a player character to play a game in a game space in which non-player characters controlled by a computer appear, the player can enjoy the game unfolding through atypical conversations between the player character and non-player character. [Brief description of the drawings]

[0015] [Figure 1] 1 is a diagram showing the main configuration of an embodiment of a game system according to the present invention; [Diagram 2] 3A to 3C are diagrams illustrating examples of characters in the game system of the present embodiment. [Diagram 3]13 shows an example of values ​​of hierarchical relationship parameters and openness parameters set for each character in the game system of this embodiment. [Figure 4] 11A to 11C are diagrams illustrating the influence of emotion parameter values ​​arising from conversations on relationship parameters in the game system of the present embodiment. [Diagram 5] 4 is an example of behavioral information in the game system of the present embodiment. [Figure 6] 6 is a flowchart illustrating the flow from the start of a game to the transition to a conversation stage in the game system of the present embodiment. [Figure 7] 6 is a flowchart illustrating the flow of a conversation stage in the game system of the present embodiment. [Figure 8] 6 is a flowchart illustrating the flow of a battle stage in the game system of the present embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0016] An embodiment of a game system according to the present invention will be described below with reference to the drawings. The game system 1 of this embodiment is for playing an action game in which a player controls a player character in a virtual space and fights against a non-player character controlled by a computer.

[0017] FIG. 1 shows the main configuration of a game system 1 of this embodiment. The game system 1 of this embodiment includes a control / processing device 10 and a player terminal 30. The control / processing device 10 of this embodiment is a cloud server. The player terminal 30 of this embodiment is a head-mounted display. The control / processing device 10 and the player terminal 30 are configured to be able to communicate with each other wirelessly or via wire. In this embodiment, a single-player game is used, but in the case of a multiplayer game, multiple player terminals 30 are provided.

[0018] The control / processing device 10 includes a storage unit 11. The storage unit 11 includes a game setting data storage unit 111, a character data storage unit 112, a software storage unit 113, a behavior information storage unit 114, and an attribute data storage unit 115.

[0019] The game setting data storage unit 111 stores display data of a virtual space in which an action game is played, display data of objects such as items used in each stage of the action game, and progress information of each stage (events in the stage and their occurrence conditions, clear conditions for the stage, etc.). The character data storage unit 112 stores display data of a plurality of characters. These characters become player characters or non-player characters depending on the selection made by the player when playing the game. The plurality of characters form pairs of two bodies. As shown in an example in FIG. 2, in this embodiment, a male character and a female character form a pair. The display data of each character also includes data of the movement (animation) of the character according to the emotion parameter. The character data storage unit 112 also stores attribute information (occupation and position, values ​​of hierarchical relationship parameters and openness parameters specific to the character, and information of the basic model) that is preset for each character. Each parameter will be described later.

[0020] The basic model is a battle style when fighting against a player character. In this embodiment, various magic spells can be cast by consuming magic points (MP) that are specified in advance. In this embodiment, four basic models are prepared: an ultimate magic type, an attack type, an enemy search type, and a defense type. The ultimate magic type is a battle style in which an orb (item) is acquired first, and when MP is sufficiently accumulated, a magic spell that consumes a large amount of MP is cast aggressively. It differs from the ultimate magic type in that a large number of magic spells that consume less MP are cast. The attack type is a battle style in which an enemy is visually identified and an attack is continued with magic as long as MP remains, and when MP runs out, an attempt is made to acquire an orb. The enemy search type is a battle style in which the enemy continues to move in the virtual space and tries to escape from the magic cast by the player character. The defense type is a battle style in which the enemy hides in the shadow of an object and attacks from a position away from the player character. Each character is preset with one of these basic models and basic abilities (values ​​such as speed and attack frequency).

[0021] The software storage unit 113 stores a conversation generation AI consisting of a trained model constructed by machine learning a large number of learning data including conversation data of conversations between people consisting of input lines and reply lines, emotion data expressing emotions arising from the conversations by the values ​​of multiple emotion parameters, and relationship data expressing relationships between people correlated with the values ​​of the emotion parameters by the values ​​of multiple relationship parameters. The conversation generation AI used in this embodiment is configured to output a reply line and the value of the emotion parameter in response to an input of an input line (or an input line and a past conversation) and a value of a relationship parameter. For such a conversation generation AI, for example, a large language model (LLM) described in Non-Patent Documents 1 and 2 can be suitably used. In this embodiment, Chat GPT is used as the conversation generation AI.

[0022] In this embodiment, eight emotion parameters are set, namely, joy, anger, sadness, enjoyment, confidence, confusion, fear, and embarrassment, and a value range from 0 to 100 is set for each of the emotion parameters. The emotion parameters of non-player characters are parameters that change each time depending on the conversation with the player character. The above eight parameters in this embodiment are merely an example, and the number and type of the parameters can be changed as appropriate. There may be only one emotion parameter.

[0023] In this embodiment, three parameters, a hierarchical relationship parameter, an openness parameter, and a reliability parameter, are set as the relationship parameters, and a value in the range of 0 to 100 is set for each of them. The hierarchical relationship parameter is a parameter indicating the authority of each character, and a value corresponding to the attributes (occupation or position) of the player character and the non-player character is set in advance. The openness parameter is a parameter indicating the degree of emotional openness of each character, and a value corresponding to the attributes (occupation or position) of the player character and the non-player character is set in advance, similar to the hierarchical relationship parameter. The higher the value of the openness parameter, the more open the character is, and the lower the value of the openness parameter, the more likely the character is to suppress his / her emotions, so to speak, wearing a mask. FIG. 3 shows examples of the values ​​of the hierarchical relationship parameter and the openness parameter set for each character.

[0024] On the other hand, the trust parameter is a parameter that indicates a non-player character's impression of the player character and changes depending on the emotions that arise during conversations with the player character. A high trust parameter value indicates a high level of intimacy or friendship with the player character, while a low trust parameter value indicates a high level of hostility toward the player character.

[0025] Emotion parameters and relationship parameters are interrelated. FIG. 4 shows the influence of the emotion parameter values ​​generated by conversation on the reliability parameter. As described above, among the relationship parameters, the hierarchical relationship parameter and the openness parameter are preset according to the attributes of the characters, and their values ​​are fixed, so they do not fluctuate depending on the emotion parameter values ​​generated by conversation. Specifically, when the player character utters an input line that increases the emotion parameter values ​​of joy, fun, or confidence, the reliability parameter value increases (indicated as "+" in FIG. 3), and when the player character utters an input line that increases the emotion parameter values ​​of anger, sadness, confusion, fear, or embarrassment, the reliability parameter value decreases (indicated as "-" in FIG. 3).

[0026] In addition, depending on the value of the relationship parameter of the non-player character, the value of the emotion parameter generated by the conversation generation IA in response to the input of the lines of the player character will differ.

[0027] The software storage unit 113 also stores text generation software that generates text data from input voice data, and text reading software that generates voice data that reads out the input text data. In this embodiment, Whisper API is used as the text generation software, and VOICEVOX (registered trademark of Kazuyuki Hiroshiba) described in Non-Patent Document 3 is used as the text reading software. Of course, these are just examples, and other software may be used.

[0028] The behavior information storing unit 114 stores behavior information in which the above-mentioned multiple emotion parameters are associated with the behavior of the non-player characters. Fig. 5 shows an example of the behavior information in this embodiment.

[0029] In this embodiment, the emotion parameters correspond to the level of the non-player character's fighting ability. Specifically, for the emotion parameters marked with "+" in Fig. 5 (joy, anger, enjoyment, confidence), the higher the parameter value, the higher the non-player character's fighting ability (speed, attack frequency, etc.), and for the emotion parameters marked with "-" in Fig. 5 (sadness, confusion, fear, embarrassment), the higher the parameter value, the lower the non-player character's fighting ability (speed, attack frequency, etc.).

[0030] The attribute data storage unit 115 stores parameter values ​​of the above-mentioned multiple emotion parameters, parameter values ​​of the above-mentioned multiple relationship parameters, and past play history (contents of conversations between player characters and non-player characters, results of battles between player characters and non-player characters, etc.).

[0031] The control and processing device 10 also includes, as functional blocks, a progress control unit 21, a display data generation unit 22, a reaction generation unit 23, and an action decision unit 24. As described above, the control and processing device 10 of this embodiment is a cloud server, and a program 20 for the control and processing device is installed in advance. These functional blocks are embodied by executing the program 20 for the control and processing device by a processor.

[0032] The player terminal 30 comprises a memory unit 31, a sound collection unit 32 and a speaker 33 provided on the sides of the head mounted display, an imaging unit 34 consisting of a plurality of cameras fixed on both sides and the top of the head mounted display, a display unit 35 located in front of the eyes when the head mounted display is worn, and an input operation unit 36. A virtual reality (VR) is displayed on the display unit 35 of the player terminal 30 of this embodiment. The input operation unit 36 ​​of this embodiment is composed of one or more buttons and a cross cursor. The input operation unit 36 ​​may have an appropriate configuration suitable for the type of player terminal 30 used and the type of game played.

[0033] The player terminal 30 also comprises, as functional blocks, a posture detection unit 41, a voice detection unit 42, a body part detection unit 43, and an input / output processing unit 44. A player terminal program 40 is pre-installed in the player terminal 30, and these functional blocks are embodied by executing the player terminal program 40 by a processor.

[0034] The posture detection unit 41 detects changes in the position and direction of the player terminal 30 (head mounted display) as changes in the posture of the head, and transmits the information to the control and processing device 10 at a predetermined cycle. The voice detection unit 42 operates only during a time period when the player is pressing a button on the input operation unit 36 ​​in a scene of conversation with a non-player character, detects the voice of the player through the sound collection unit 32, and transmits the voice data to the control and processing device 10. The voice detection unit 42 may detect voice regardless of whether or not there is an operation on the input operation unit 36, but in that case, the processing load related to the detection of voice will be large, so in this embodiment, it is configured to operate only during a time period when the button on the input operation unit 36 ​​is pressed. The body part detection unit 43 detects the positions of predetermined body parts (fingertips, arm joints, toes, foot joints, etc.) of the player from images captured at a predetermined cycle by multiple cameras of the imaging unit 34, and transmits the positions to the control and processing device 10 at a predetermined cycle (for example, frame rate). The input / output processing unit 44 displays image data of the virtual space received from the control / processing device 10 on the display unit 35. In addition to the function of displaying images on the display unit 35, the input / output processing unit 44 also reads input operations made by the player on the screen and transmits the information to the control / processing device 10.

[0035] Next, the flow of playing a game in the game system 1 of this embodiment will be described with reference to FIGS.

[0036] When the user starts up the player terminal 30, a game start screen (title screen) is displayed on the display unit 35. When the user presses the start button displayed on the game start screen (title screen) by a predetermined operation, this information is transmitted to the control and processing device 10.

[0037] During game play, in the control / processing device 10, the display data generation unit 22 generates display data for the virtual space at the start of the game and transmits it to the player terminal 30. In the player terminal 30, the input / output processing unit 44 causes the received data to be displayed on the display unit 35. In addition, the progress control unit 21 develops the game according to a previously prepared flow, and the display data generation unit 22 sequentially generates display data for the virtual space according to the development, and display data for the player characters and non-player characters, and transmits it to the player terminal 30. In the player terminal 30, the input / output processing unit 44 causes the display unit 35 to sequentially display the received data.

[0038] In this embodiment, when the game is started, the progress control unit 21 reads out information on a plurality of characters (appearances and attribute data of the player characters) stored in the character data storage unit 112, and transmits it to the player terminal 30. In the player terminal 30, the input / output processing unit 44 displays the appearances of the plurality of characters on the display unit 35 based on the received data (step 1). When the player selects one of the characters as a player character (PC) (step 2), the character is determined as the player character. In addition, the character paired with the character selected by the player is determined as a non-player character (NPC) (step 3). Information on the determined player character and non-player character is transmitted to the control / processing device 10, and the progress control unit 21 stores the information in the attribute data storage unit 115. In addition, the progress control unit 21 reads out data on the character from the character data storage unit 112 and stores it in the attribute data storage unit 115 (step 4). Note that since the player himself operates the player character, various data (attributes, relationship parameters, basic model, etc.) previously set for the character selected as the player character are not used.

[0039] Once the player character and non-player character have been determined, a screen is displayed in which the player character and non-player character converse with each other. If the player is playing the game for the first time (YES in step 11), the non-player character starts the conversation by asking a prepared question about the player character's gender, age, hobbies, etc. (step 12). The questions from the non-player character are prepared in advance as text data and voice data, which are transmitted to the player terminal 30, the text data being displayed on the screen of the display unit 35, and the voice data being output from the speaker 33. When the player presses a button on the input operation unit 36 ​​to answer the question by voice (step 14), the voice detection unit 42 detects the voice of the player through the sound collection unit 32 and transmits the voice data to the control and processing device 10.

[0040] In the control / processing device 10, the reaction generation unit 23 inputs the player's voice received from the player terminal 30 into text generation software to generate text data. The text data of the lines of the player character is instantly transmitted to the player terminal 30 and displayed on the display unit 35 by the input / output processing unit 44.

[0041] Here, since the player is playing the game for the first time, the attribute data storage unit 115 stores only the attributes of the player character and the non-player character, and the values ​​of the hierarchical relationship parameter and openness parameter corresponding to the attributes, and no data related to the conversation is stored. The reaction generation unit 23 inputs the lines of the player character, the initial values ​​of the emotion parameter, the values ​​of the hierarchical relationship parameter and openness parameter (these are fixed values), and the initial values ​​of the relationship parameter to the conversation generation AI (step 15). The initial values ​​of the emotion parameter and the reliability parameter are both, for example, 0. The emotion parameter values ​​being 0 means that no emotion is given to the non-player character. Also, the reliability parameter value being 0 means that the non-player character has no impression of the player character (i.e., it is the first time they meet). Here, the emotion parameter values ​​and the reliability parameter values ​​are both set to 0 when playing the game for the first time, but instead, random variables may be generated and used as the initial values ​​of the parameters, and emotions and impressions of the player character at the first meeting may be given.

[0042] When the conversation based on the prepared standard questions ends, the reaction generation unit 23 stores the content of the player's answer in the attribute data storage unit 115. Thereafter, whenever the voice detection unit 42 detects that the player has spoken (step 14) and the voice data of the player is input to the control and processing device 10, the voice data of the player is input to the text generation software to generate text data, and the text data (the player character's lines; input lines), past conversation data stored in the attribute data storage unit 115, and the relationship parameter values ​​at that time are input to the conversation generation AI (step 15), and the text data of the player character's lines, the non-player character's reply lines (reply lines), and the emotion parameter values ​​are output (step 16). Note that, since the values ​​of the hierarchical relationship parameter and the openness parameter among the relationship parameters are fixed values, these values ​​may be input only the first time and the conversation generation AI may continue to hold those values.

[0043] The reaction generation unit 23 further updates the value of the reliability parameter stored in the attribute data storage unit 115 using an algorithm based on the output emotion parameter value and the relationship described with reference to FIG. 4 (step 17). The conversation data, emotion parameters, and relationship parameters thus generated are sequentially stored in the attribute data storage unit 115 (step 18). The reaction generation unit 23 inputs the text data of the reply into the text-to-speech software, which generates voice data of the non-player character. Furthermore, the display data generation unit 22 reads out from the character data storage unit 112 the action data of the non-player character that corresponds to the emotion parameter with the highest value among the multiple emotion parameters at that time.

[0044] Once the text data and voice data of the reply lines have been generated as described above and the action data of the non-player characters has been read out, the reaction generation unit 23 transmits these data to the player terminal 30. In the player terminal 30, the input / output processing unit 44 displays the action data of the non-player characters as animation on the screen of the display unit 35, and displays the text of the reply lines adjacent to it. Furthermore, the input / output processing unit 44 outputs the voice data of the reply lines of the non-player characters from the speaker 33.

[0045] If the player has played the game in the past (NO in step 11), after the game starts, the reaction generation unit 23 reads out past conversation data stored in the attribute data storage unit 115 (step 13), inputs it to the conversation generation AI together with the relationship parameter value (step 15), and generates the lines (reply lines) of the non-player character and the value of the emotion parameter (step 16). In addition, the same process as above is used to generate text data of the lines of the player character and a new reliability parameter value. Note that the past conversations stored in the attribute data storage unit 115 do not necessarily need to be stored as sentences, and may be stored as a group of words (tokens) extracted from sentences, or the like.

[0046] If all past conversation data stored in the attribute data storage unit 115 is input to the learned model every time, it may take time to generate a reply line, etc. In that case, it is preferable to input only the most recent predetermined number of conversations to the conversation generation AI, or to input the contents of past conversations to the conversation generation AI only when playing the game, and thereafter input only the lines of the player character, the parameter values ​​of the emotion parameters, and the parameter values ​​of the relationship parameters. In addition, if it takes time for the conversation generation AI to generate a reply line, etc., the display data generation unit 22 may generate display data that causes the non-player character to act as if he or she is deep in thought, and display the display data on the display unit 35 during the period from when the player utters a sound until the conversation generation AI generates a reply line, etc.

[0047] Not only when a conversation starts, but also when the conversation is interrupted, the content of the conversation up to that point and the parameter values ​​of the relationship parameters stored in the attribute data storage unit 115 are input to the conversation generation AI model in the same manner as described above to generate lines for the non-player characters. For example, if a conversation was held in the previous play about a school test, a line asking about the results of the test is generated.

[0048] When the player moves his / her body while talking with a non-player character, the posture detection unit 41 detects the change in position and orientation of the player terminal 30 (head mounted display) as a change in the posture of the head, and the body part detection unit 43 detects the positions of predetermined body parts of the player (fingertips, arm joints, toes, foot joints, etc.). The detected data is transmitted to the control / processing device 10 at a predetermined cycle (for example, frame rate). In the control / processing device 10, the display data generation unit 22 generates display data representing the movement of the player character based on the received data, and transmits it to the player terminal 30. In the player terminal 30, the input / output processing unit 44 displays the received display data on the display unit 35.

[0049] When conversations between the player character and non-player character occur a predetermined number of times (e.g., 30 times) (YES in step 19), the progress control unit 21 invites the player character into the battle stage. Here, it is after conversations have occurred a predetermined number of times, but the player character may be invited into the battle stage in response to a request from the player to start a battle before a predetermined time has elapsed.

[0050] When moving to the battle stage, the progress control unit 21 causes the conversation generation AI to generate lines for the non-player characters to inform the player that the battle stage has begun. For example, a line such as "It's time for battle" is generated (step 20). At this time, data on past battle results stored in the attribute data storage unit 115 may also be input to the conversation generation AI to generate lines for the non-player characters (for example, lines based on the results of the previous battle such as "I won't lose this time").

[0051] The display data generating unit 22 generates display data for the battle stage and displays it on the display unit 35. The battle stage in this embodiment is a battle space provided in a virtual space, in which objects in which a character can hide and items such as orbs that a character can use to cast magic are placed.

[0052] When moving to the battle stage, the progress control unit 21 reads out the basic model and fighting ability of the non-player character stored in the character data storage unit 112 (step 21). The action decision unit 24 also reads out the value of the emotion parameter stored in the attribute data storage unit 115 at that time. Then, the action decision unit 24 increases or decreases the fighting ability of the non-player character from its initial value based on the action information stored in the action information storage unit 114 (see FIG. 5, step 22). Here, the fighting ability is increased or decreased from its initial value based on the value of the emotion parameter of the non-player character, but in addition to this, the fighting ability may be further increased or decreased based on the result of the previous battle (if the player character won the previous battle, the fighting ability may be increased, and if the player character lost, the fighting ability may be decreased).

[0053] Once the fighting abilities of the non-player characters have been determined, a battle start button is displayed on the display unit 35. When the user presses this button, the battle begins.

[0054] In the battle stage of this embodiment, the player character moves freely in the battle space based on the abilities (hit points, magic points, movement speed, etc.) that are provided to the player character in advance, while activating a type of magic that is set to the player character in advance to hit a non-player character. Magic points (MP) are needed to activate magic. Orbs (items) for recovering MP are placed in the battle space, and the player character and non-player character obtain the orbs while dodging the opponent's magic, and activate magic using the MP recovered thereby. When the player character's magic hits a non-player character or the non-player character's magic hits a player character (YES in step 24), the battle ends (step 25).

[0055] When the player moves or changes the direction of his / her head while playing the battle stage, the change in the player's position and the change in the head's posture are detected by the posture detection unit 41. Also, when the player moves his / her limbs, the movement is detected by the body part detection unit 43 via the imaging unit 34. This information is sequentially transmitted to the control / processing device 10. In the control / processing device 10, the display data generation unit 22 changes the position and the orientation of the limbs of the player character in the virtual space based on the received data, and also sets the viewpoint of the player character in the virtual space according to the posture of the player's head.

[0056] When the battle stage ends, the progress control unit 21 creates a battle result (step 26). The battle result includes information such as the battle time, the movement distance, and the number of attacks. For example, the shorter the battle time, the greater the difference in ability between the player character and the non-player character is determined to be. Also, for example, if the movement distance is long and the number of attacks is small, it is determined that the player character mainly fled the battle space. The reaction generation unit 23 inputs the battle result thus generated to the conversation generation AI together with the relationship data stored in the attribute data storage unit 115 (step 27), and generates lines according to the battle result (step 28). At this time, the battle result is put into a conversation-like template prepared in advance and input to the conversation generation AI, which interprets this as one input line and generates a reply line. For example, if the player character wins, a line such as "You were strong, I'm sorry" is generated, and if the non-player character wins, a line such as "That was close." is generated.

[0057] As described above, in this embodiment, one play ends with a conversation with a non-player character and a battle. When the play ends, the content of the conversation between the player character and the non-player character in that play, the relationship parameter value of the non-player character at that time, and the battle result are stored in the attribute data storage unit 115.

[0058] As described above, in the game system 1 of this embodiment, any voice uttered by the player (the lines of the player character) is input to the conversation generation AI, and a reply line corresponding to the line is generated. In addition, different values ​​of the emotion parameters of the player character are generated according to the conversation between the player character and the non-player character, and the value of the reliability parameter varies depending on the value of the emotion parameter. Then, the fighting ability of the non-player character is determined according to the value of the emotion parameter at the time of transition to the battle stage. Therefore, it is possible to enjoy atypical conversations and game developments between the player character and the non-player character.

[0059] The above embodiment is merely an example and can be modified as appropriate in accordance with the spirit of the present invention.

[0060] In the above embodiment, a conversation is made between the player character and a non-player character who is an opponent, but a configuration in which a conversation is made between the player character and a non-player character who is a companion character who fights against an enemy character together with the player character can also be adopted. In addition, the configuration is not limited to battle-type games, and can be configured to play any kind of game in which a non-player character may appear, such as a role-playing game.

[0061] In the above embodiment, there is one non-player character, but there may be more than one non-player character. In that case, the action of each non-player character may be determined based on the value of an emotion parameter that represents the emotion that arises in a conversation with a player character. Also, the game may be a so-called multiplayer game in which multiple player characters exist.

[0062] In the above embodiment, a head mounted display was used as the player terminal 30, but the player terminal 30 may be any terminal having a function for acquiring the voice of the player and a function for displaying the space in which the game is played, and for example, the player terminal may be configured with a microphone and a display. Also, in the above embodiment, a virtual space was displayed as the game space on the display unit 35, but it is also possible to use a device that displays augmented reality as the game space, or a display that displays a general two-dimensional screen.

[0063] In the above embodiment, the conversation generation AI generates a response line and an emotion parameter value in response to the input of the player character's line and the relationship parameter value, and updates the reliability parameter value using the generated emotion parameter value and a previously prepared algorithm, but a conversation generation AI that also generates a relationship parameter value may be used. Also, the emotion parameter value may be input to the conversation generation AI, and the conversation generation AI may update the emotion parameter value. The content to be input to the conversation generation AI and the content to be generated may be designed according to the characteristics of the game, and a learned model may be constructed by having the learning model perform machine learning using learning data according to the design.

[0064] In the above embodiment, the game system is configured with a cloud server as the control / processing device 10 and a head-mounted display as the player terminal 30, taking into consideration that a large GPU resource is required for processing in the LLM. However, when a head-mounted display equipped with a large GPU resource or the like can be used, each memory unit and each functional block provided in the control / processing device 10 may be provided in the player terminal 30, and the game system may be configured compactly and integrally with only the player terminal 30. Also, some of the memory units and functional blocks provided in the control / processing device 10 may be provided in the player terminal 30. For example, only the software memory unit 113 and the attribute data memory unit 115 of the memory unit 11, and only the reaction generation unit 23 of the functional blocks may be provided in the control / processing device 10, and the other memory units and functional blocks provided in the control / processing device 10 in the above embodiment may be provided in the player terminal 30.

[0065] In the above embodiment, the player inputs the lines of the player character as voice and outputs the replies of the non-player characters as voice, but this is not essential to the present invention. If the player inputs the lines of the player character as text and outputs the replies of the non-player characters as text, it is not necessary to use text generation software and text-to-speech software.

[0066] [Aspects] It will be apparent to those skilled in the art that the above-described exemplary embodiments are illustrative of the following aspects.

[0067] (Section 1) One aspect of the present invention is a system for allowing a player to control a player character to play a game in a game space in which a non-player character controlled by a computer appears, the system comprising: a display data storage unit in which display data of the game space, display data of the player character, and display data of the non-player character are stored; a trained model created by machine learning learning data including conversation data of a conversation consisting of an input line and a reply line, and emotion data expressing emotions arising from the conversation by emotion parameter values, the trained model storing a trained model that generates a reply line and the emotion parameter values ​​in response to an input of an input line; a behavior information storage unit in which behavior information that associates the emotion parameter value with the behavior of the non-player character is stored; a dialogue input unit for inputting dialogue of the player character; a reaction generation unit that inputs the lines of the player character input to the line input unit as input lines to the trained model to generate a response line and a value of the emotion parameter; a dialogue output unit that outputs the response dialogue generated by the trained model by the reaction generation unit as a dialogue of the non-player character; a behavior determination unit that determines a behavior of the non-player character in accordance with the value of the emotion parameter generated in the trained model by the reaction generation unit in light of the behavior information stored in the behavior information storage unit, and causes the non-player character to act in the game space; The present invention is characterized by comprising:

[0068] In the game system according to paragraph 1, the player inputs the lines of the player character that the player operates into the line input unit. This input may be text input or voice input. When the lines of the player character are input, the reaction generation unit inputs the input lines into the trained model to generate a reply line and an emotional parameter value. The generated lines are output by the line output unit as the lines of the non-player character. This output may also be text output or voice output. In addition, the behavior determination unit determines the behavior of the non-player character based on the generated emotional parameter value in light of previously prepared behavior information, and causes the non-player character to operate in the virtual space.

[0069] In the game system according to paragraph 1, a response line is generated in response to any line spoken by the player character, so that a non-standard conversation can be enjoyed between the player character and the non-player character. Also, an emotion parameter value is generated in response to the line spoken by the player character, and the action of the non-player character is determined in response to the value, so that a different game development can be enjoyed depending on the line spoken by the player character.

[0070] (Section 2) The game system according to paragraph 2 is the game system according to paragraph 1, The game is a game in which the player character and the non-player character compete against each other, The behavior information includes information that associates the value of the emotion parameter with the level of ability of the non-player character.

[0071] In the game system according to paragraph 2, the abilities and battle style of the non-player characters during battle change according to the lines of the player character, allowing players to enjoy a variety of game developments.

[0072] (Section 3) The game system according to paragraph 3 is a game system according to either paragraph 1 or paragraph 2, A plurality of emotion parameters are set.

[0073] In the game system according to paragraph 3, by using a plurality of emotion parameters, the actions of the non-player characters can be set in a more complex manner, and a more diverse game development can be enjoyed.

[0074] (Section 4) The game system according to paragraph 4 is the game system according to paragraph 3, a relationship parameter that changes depending on a part or all of the values ​​of the plurality of emotion parameters is set; The reaction generation unit inputs the input line and the value of the relationship parameter to the trained model to generate the reply line and the value of the emotion parameter.

[0075] In the game system according to paragraph 4, relationship parameters change according to the values ​​of multiple emotion parameters that arise during conversations between the player character and non-player characters, so that the longer the game is played, the deeper the relationship with the non-player character can be.

[0076] (Section 5) The game system according to paragraph 5 is a game system according to any one of paragraphs 1 to 4, further comprising: An attribute data storage unit for storing past conversations between the player character and the non-player character. Equipped with Past conversations stored in the attribute data storage unit are further input to the trained model to generate the response lines and the emotion parameter values.

[0077] In the game system according to paragraph 5, reply lines can be generated that take past conversations into consideration.

[0078] (Section 6) The game system according to paragraph 6 is the game system according to paragraph 5, When no lines of the player character are input to the line input unit for a predetermined time, the reaction generation unit inputs the past conversation into the trained model to generate the response lines.

[0079] In the game system according to paragraph 6, when no lines are input for the player character, a non-player character can utter a line to start a conversation. [Explanation of symbols]

[0080] 1. Game System 10...Control and processing device 11...Storage section 111...Game setting data storage unit 112…Character data storage unit 113...Software storage unit 114…Behavior information storage unit 115...attribute data storage unit 20...Program for control and processing device 21...Progress control section 22...Display data generation unit 23...Reaction product section 24…Action Decision-Making Department 30…Player terminal 31...Storage section 32…Sound collection section 33…Speaker 34…Imaging unit 35...Display section 40...Player terminal program 41...Attitude detection unit 42...Voice detection section 43...Part detection unit 44... Input / output processing section

Claims

1. A system for allowing a player to control a player character to play a game in a game space in which non-player characters controlled by a computer appear, comprising: a display data storage unit in which display data of the game space, display data of the player character, and display data of the non-player character are stored; a trained model created by machine learning learning data including conversation data of a conversation consisting of an input line and a reply line, and emotion data expressing emotions arising from the conversation by emotion parameter values, the trained model storing a trained model that generates a reply line and the emotion parameter values ​​in response to an input of an input line; a behavior information storage unit in which behavior information that associates the emotion parameter value with the behavior of the non-player character is stored; a dialogue input unit for inputting dialogue of the player character; a reaction generation unit that inputs the lines of the player character input to the line input unit as input lines to the trained model to generate a response line and a value of the emotion parameter; a dialogue output unit that outputs the response dialogue generated by the trained model by the reaction generation unit as a dialogue of the non-player character; a behavior determination unit that determines a behavior of the non-player character in accordance with the value of the emotion parameter generated in the trained model by the reaction generation unit in light of the behavior information stored in the behavior information storage unit, and causes the non-player character to move in the game space; A game system comprising:

2. The game is a game in which the player character and the non-player character compete against each other, The behavior information includes information associating the value of the emotion parameter with the level of the ability of the non-player character.

2. The game system according to claim 1 .

3. A plurality of the emotion parameters are set.

2. The game system according to claim 1 .

4. a relationship parameter is set that changes depending on a value of a part or all of the plurality of emotion parameters, The reaction generation unit inputs the input line and the value of the relationship parameter to the trained model to generate the reply line and the value of the emotion parameter.

4. The game system according to claim 3.

5. moreover, An attribute data storage unit for storing past conversations between the player character and the non-player character. It is equipped with The learned model further inputs past conversations stored in the attribute data storage unit to generate the response lines and the emotion parameter values.

2. The game system according to claim 1 .

6. The reaction generation unit, when no lines of the player character are input to the line input unit for a predetermined time, inputs the past conversation into the trained model to generate the reply lines.

6. The game system according to claim 5.

Images