Conversation processing method, conversation processing device, program product and electronic device

Abstract

The invention provides a dialogue processing method, a dialogue processing device, a program product and electronic equipment, and relates to the technical field of computers. The method comprises the following steps: acquiring background information of a plurality of non-user virtual characters, and generating first dialogue information among the plurality of non-user virtual characters according to the background information of the plurality of non-user virtual characters; obtaining second dialogue information input by the first user or the first user virtual role in the dialogue group; the first user virtual role is a virtual role controlled by the first user; the dialogue group comprises the plurality of non-user virtual roles; identifying a dialogue object of the second dialogue information; generating third dialogue information in the dialogue group according to the second dialogue information and the dialogue object; the speaker of the third dialogue information is the non-user virtual character. Diversified dialogue information of a non-user virtual role can be generated, and the user experience is improved.

Description

Technical Field

[0001] The present disclosure relates to the field of computer technologies, and in particular, to a dialogue processing method, a dialogue processing apparatus, a computer program product, and an electronic device. Background Art

[0002] In scenarios such as games and virtual social platforms, non-user virtual characters are usually set, which refer to virtual characters controlled by a computer rather than a user. A user can have a dialogue with a non-user virtual character.

[0003] In related technologies, the dialogue information of non-user virtual characters is generally relatively fixed. For example, most of the fixed lines are pre-written by production staff. When a user has a dialogue with a non-user virtual character, the dialogue content is presented according to the lines. Such a dialogue effect is relatively monotonous, affecting the user experience. Summary of the Invention

[0004] The present disclosure provides a dialogue processing method, a dialogue processing apparatus, a computer program product, and an electronic device, so as to at least to some extent solve the problem that the dialogue information of non-user virtual characters is relatively fixed.

[0005] According to a first aspect of the present disclosure, there is provided a dialogue processing method, the method including: obtaining background information of a plurality of non-user virtual characters, and generating first dialogue information among the plurality of non-user virtual characters according to the background information of the plurality of non-user virtual characters; obtaining second dialogue information input by a first user or a first user virtual character in a dialogue group; the first user virtual character is a virtual character controlled by the first user; the dialogue group includes the plurality of non-user virtual characters; identifying a dialogue object of the second dialogue information; generating third dialogue information in the dialogue group according to the second dialogue information and the dialogue object; the speaker of the third dialogue information is the non-user virtual character.

[0006] According to a second aspect of the present disclosure, there is provided a dialogue processing apparatus, the apparatus including: a first dialogue generation module configured to obtain background information of a plurality of non-user virtual characters, and generate first dialogue information among the plurality of non-user virtual characters according to the background information of the plurality of non-user virtual characters; a second dialogue obtaining module configured to obtain second dialogue information input by a first user or a first user virtual character in a dialogue group; the first user virtual character is a virtual character controlled by the first user; the dialogue group includes the plurality of non-user virtual characters; a dialogue object identification module configured to identify a dialogue object of the second dialogue information; a third dialogue generation module configured to generate third dialogue information in the dialogue group according to the second dialogue information and the dialogue object; the speaker of the third dialogue information is the non-user virtual character.

[0007] According to a third aspect of the present disclosure, there is provided a computer program product including a computer program which, when executed by a processor, implements the method and its possible implementations of the first aspect described above.

[0008] According to a fourth aspect of the present disclosure, there is provided an electronic device including: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to execute the executable instructions to execute the method and its possible implementations of the first aspect described above.

[0009] The technical solution of the present disclosure has the following beneficial effects:

[0010] Generate first dialogue information between multiple non-user virtual characters according to the background information of the multiple non-user virtual characters, obtain second dialogue information input by a first user or a first user virtual character in a conversation group, identify the conversation object of the second dialogue information, and generate third dialogue information according to the second dialogue information and the conversation object. An implementation solution for group chat between a user and multiple non-user virtual characters is provided. The first dialogue information is introduced, and third dialogue information of non-user virtual characters is generated for the second dialogue information and the conversation object of the second dialogue information, improving the diversity and interest of the dialogue information of non-user virtual characters, improving the problem of relatively fixed dialogue information, and enhancing the user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 Shows a system architecture diagram in this exemplary embodiment;

[0012] Figure 2 Shows a flowchart of a dialogue processing method in this exemplary embodiment;

[0013] Figure 3 Shows a flowchart of generating first dialogue information in this exemplary embodiment;

[0014] Figure 4 Shows a flowchart of generating third dialogue information in this exemplary embodiment;

[0015] Figure 5 Shows a schematic flowchart of a dialogue processing method in this exemplary embodiment;

[0016] Figure 6 Shows a structural schematic diagram of a dialogue processing device in this exemplary embodiment;

[0017] Figure 7 Shows a structural schematic diagram of an electronic device in this exemplary embodiment. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] The exemplary embodiments of the present disclosure will be described more fully hereinafter with reference to the accompanying drawings.

[0019] The accompanying drawings are schematic illustrations of the present disclosure and are not necessarily drawn to scale. Some of the block diagrams shown in the accompanying drawings may be functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities may be implemented in software, or in a hardware module or integrated circuit, or in a network, processor, or microcontroller. The embodiments can be implemented in various forms and should not be construed as limited to the examples set forth herein. The features, structures, or characteristics described in the present disclosure may be combined in any suitable manner in one or more embodiments. In the following description, numerous specific details are provided to give a full understanding of the embodiments of the present disclosure. However, those skilled in the art will realize that one or more of the specific details may be omitted in implementing the technical solutions of the present disclosure, or other methods, components, devices, steps, etc. may be used to replace one or more specific details.

[0020] In the related art, the dialogue information of non-user virtual characters is generally relatively fixed. For example, most of the lines are pre-written by production staff and usually only have a fixed few sentences or dozens of sentences. When a user converses with a non-user virtual character, the dialogue content is presented according to the lines. Such dialogue effects are relatively monotonous, resulting in a reduction in the sense of reality in scenarios such as games and virtual social platforms, and affecting the user experience.

[0021] In view of the above problems, the exemplary embodiments of the present disclosure provide a dialogue processing method that can improve the problem of relatively fixed dialogue information of non-user virtual characters.

[0022] Figure 1The system architecture diagram of the operating environment of this exemplary embodiment is shown. The system architecture may include a terminal device 110 and a server 120. Among them, the terminal device 110 may be a mobile phone, a tablet computer, a personal computer, an intelligent wearable device, a game console, and other devices. Applications such as game programs and virtual social programs are installed on the terminal device 110. In one embodiment, the terminal device 110 has a display function and can display a graphical user interface. The graphical user interface may include an interface of an operating system or an interface of an application program, and the terminal device 110 may display conversation information in the graphical user interface. In one embodiment, the terminal device 110 has an audio function and can play conversation information in an audio manner. The server 120 generally refers to a background system that provides services such as games and virtual social networking in this exemplary embodiment, and may be a server or a cluster of multiple servers. In one embodiment, a game server program is deployed on the server 120 for executing game data processing on the server side. The terminal device 110 and the server 120 may be connected via a wired or wireless communication link to perform data transmission. The method in this exemplary embodiment may be executed by any one or more of the terminal device 110 and the server 120.

[0023] In one embodiment, the above method can be implemented and executed based on a cloud interactive system. The cloud interactive system can be the above system architecture. Various cloud applications can be run under the cloud interactive system, such as cloud games. Taking cloud games as an example, cloud games refer to a game method based on cloud computing. In the operation mode of cloud games, the operation subject of the game program and the game screen presentation subject are separated, and the storage and operation of the control and interaction methods in the game are completed on the cloud game server (such as the above server 120), and the cloud game client (such as the above terminal device 110) is used for receiving and sending data and presenting the game screen. For example, the cloud game client can be a display device with data transmission function close to the user side, such as a mobile terminal, a TV, a computer, a handheld computer, etc.; and the cloud game server in the cloud performs information processing. When playing the game, the user operates the cloud game client to send an operation instruction to the cloud game server. The cloud game server runs the game according to the operation instruction, encodes and compresses the game screen and other data, and returns it to the cloud game client through the network. Finally, the cloud game client decodes and outputs the game screen.

[0024] In one implementation, the above method can be implemented by the terminal device 110 without deploying the server 120. For example, in a stand-alone environment, a stand-alone application is installed on the terminal device 110 to execute the above method.

[0025] Figure 2 An exemplary process of a dialog processing method is shown, which may include the following steps:

[0026] Step S210: Obtain the background information of multiple non-user virtual characters, and generate first dialogue information among the multiple non-user virtual characters according to the background information of the multiple non-user virtual characters;

[0027] Step S220: Obtain second dialogue information input by the first user or the first user virtual character in the dialogue group; the first user virtual character is a virtual character controlled by the first user; the dialogue group includes the above-mentioned multiple non-user virtual characters;

[0028] Step S230: Identify the dialogue object of the second dialogue information;

[0029] Step S240: Generate third dialogue information in the dialogue group according to the second dialogue information and the dialogue object; the speaker of the third dialogue information is a non-user virtual character.

[0030] Based on Figure 2 the method shown above, generate first dialogue information among multiple non-user virtual characters according to the background information of the multiple non-user virtual characters, obtain second dialogue information input by the first user or the first user virtual character in the dialogue group, identify the dialogue object of the second dialogue information, and generate third dialogue information according to the second dialogue information and the dialogue object. A solution for group chat between a user and multiple non-user virtual characters is provided. By introducing the first dialogue information and generating third dialogue information of non-user virtual characters for the second dialogue information and the dialogue object of the second dialogue information, the diversity and interest of the dialogue information of non-user virtual characters are improved, the problem of relatively fixed dialogue information is solved, and the user experience is enhanced.

[0031] Next, Figure 2 each step will be described in detail.

[0032] Referring to Figure 2 , in step S210, obtain the background information of multiple non-user virtual characters, and generate first dialogue information among the multiple non-user virtual characters according to the background information of the multiple non-user virtual characters.

[0033] A non-user virtual character refers to a virtual character controlled by a computer rather than a user, such as a non-player character (NPC) in a game. In contrast, a user virtual character refers to a virtual character controlled by a user, such as a player character in a game.

[0034] In scenarios such as games and virtual social platforms, background information can be set for non-user virtual characters, for example, by production staff. The background information of non-user virtual characters includes, but is not limited to: age, occupation, gender, family information, personality, etc. First dialogue information between non-user virtual characters is generated based on the background information of the non-user virtual characters, and this dialogue information can be generated without user participation.

[0035] In one implementation, as shown in Figure 3 the above-mentioned generation of the first dialogue information between multiple non-user virtual characters based on the background information of the multiple non-user virtual characters may include the following steps S310 to S330:

[0036] Step S310, determine the dialogue topic.

[0037] This dialogue topic is the dialogue topic of the first dialogue information. Exemplarily, the dialogue topic can be specified by a user (such as a first user). Alternatively, the dialogue topic can be automatically determined by the program. For example, multiple dialogue topics are pre-set, and the program selects one of them (such as randomly or sequentially) as the dialogue topic of the first dialogue information.

[0038] In one implementation, the non-user virtual character is a non-player character in a game. The above-mentioned determination of the dialogue topic may include the following steps:

[0039] Determine the first prompt information according to the background information of the game;

[0040] Input the first prompt information into the first generation model, and generate the dialogue topic through the first generation model.

[0041] Among them, the background information of the game includes, but is not limited to: the environment, era, and location where the game story takes place; the world view setting of the game; relevant history, culture, social structure, etc.; the background information of the main characters in the game.

[0042] A generative model refers to a generative machine learning model. The generative models of the present disclosure can be generative models for natural language processing, such as LLM (Large Language Model), RNN (Recurrent Neural Network), GAN (Generative Adversarial Network), Transformer, BERT (Bidirectional Encoder Representations from Transformers), etc. Embodiments of the present disclosure relate to one or more generative models, including a first generative model, a second generative model, a third generative model, a fourth generative model, a fifth generative model, etc. These generative models can be different generative models, including differences in any one or more aspects such as type, structure, parameters, etc. For example, the first generative model, the second generative model, the third generative model, the fourth generative model, and the fifth generative model are all LLMs, but with different parameters. Or, any two or more generative models can be the same model. For example, the first generative model, the second generative model, the third generative model, the fourth generative model, and the fifth generative model are all the same LLM.

[0043] A prompt is information used to prompt a generative model, usually describing the characteristics of the content to be generated, so that the generative model can generate content that conforms to the prompt under the guidance of the prompt. The prompt can be obtained by combining a prompt template and key information. For example, for different generative tasks, corresponding prompt templates can be preset in advance. The prompt template includes one or more positions where key information can be filled in, and the key information is filled in these positions to obtain the corresponding prompt.

[0044] In the embodiments of the present disclosure, the first prompt is the prompt for generating a conversation topic, and the first generative model is the generative model for generating a conversation topic. Exemplarily, a first prompt template can be obtained, and the background information of the game is combined with the first prompt template to obtain the first prompt.

[0045] The first prompt template is as follows:

[0046] "# Task

[0047] I want you to act as a talk show comedian.

[0048] Actor information: {NPC information}.

[0049] The world where the actor is located: {the background information of the game}.

[0050] You need to use your wisdom, creativity and observation skills to generate 20 talk show topics.

[0051] Talk show topics such as "How to humorously view weight loss".

[0052] You should ensure that personal anecdotes or experiences of daily activities are incorporated into the topics to make them more relevant and appealing to the audience.

[0053] Please output in XML format:<Topic>[Topic 1]< / Topic><Topic>[Topic 2]< / Topic>...<Topic>[Call 20]< / Topic>”

[0054] Fill in the NPC information in the game and the background information of the game into the above template to obtain the first prompt message. Input the first prompt message into the first generation model to obtain the conversation topic (i.e., the talk show topic).

[0055] Step S320, determine the second prompt message according to the conversation topic and the background information of multiple non-user virtual characters.

[0056] Step S330, input the second prompt message into the second generation model, and generate the first conversation message through the second generation model.

[0057] Among them, the second prompt message is the prompt message for generating the first conversation message, and the second generation model is the generation model for generating the first conversation message. Exemplarily, a second prompt template can be obtained, and the conversation topic and the background information of multiple non-user virtual characters are combined with the second prompt template to obtain the second prompt message.

[0058] The second prompt template is as follows:

[0059] "I want you to play multiple talk show comedians, and the relationship is strangers.

[0060] The information of actor 1 is:

[0061] {NPC1 information}

[0062] The information of actor 2 is:

[0063] {NPC2 information} ...

[0065] The information of actor n is:

[0066] {NPCn information}

[0067] I will provide you with a topic related to current events. You will use your wisdom, creativity and observation skills to create a talk show conversation between strangers based on these topics. The conversation should be concise and not have too much logic.

[0068] My request is "{Talk show theme}".

[0069] You should ensure to incorporate personal anecdotes or experiences into daily activities to make it more relevant and appealing to the audience.

[0070] Due to the relationship of being strangers, an opening statement about the occasion of meeting is needed.

[0071] The output format is:

[0072] <Dialogue>

[0073] Actor 1:...

[0074] Actor 2:...

[0075] Actor 1:...

[0076] Actor 2:... ...

[0078] < / Dialogue>

[0079] Fill in the dialogue topic (i.e., the talk show theme), background information of multiple non-user virtual characters (i.e., NPC1 information, NPC2 information, etc.) into the above template to obtain the second prompt information. Input the second prompt information into the second generation model to obtain the first dialogue information.

[0080] Based on Figure 3 the method, generating the first dialogue information that matches the dialogue topic and background information of non-user virtual characters can improve the diversity and interest of the first dialogue information and attract users.

[0081] In one implementation, the non-user virtual character is a non-player character in the game. The dialogue processing method may further include the following steps:

[0082] Determine the third prompt information according to the background information of the game and the first dialogue information;

[0083] Input the third prompt information into the third generation model, and generate the modified first dialogue information through the third generation model.

[0084] Among them, the third prompt information is the prompt information for modifying the first dialogue information, and the third generation model is the generation model for modifying the first dialogue information. Exemplarily, a third prompt template can be obtained, and the background information of the game and the first dialogue information are combined with the third prompt template to obtain the third prompt information.

[0085] The third prompt template is as follows:

[0086] "Rewrite the following conversation into a conversation that conforms to the game world among multiple NPCs.

[0087] Game world:

[0088] {Background information of the game}

[0089] The information of NPC1 is:

[0090] {NPC1 information}

[0091] The information of NPC2 is:

[0092] {NPC2 information} ...

[0094] The personality of NPCn is:

[0095] {NPCn information}

[0096] When rewriting, only the dialogue text is required, without other content such as actions and expressions.

[0097] {Dialogue}"

[0098] Fill the first dialogue information, the background information of the game, and the background information of multiple non-user virtual characters (i.e., NPC1 information, NPC2 information, etc.) into the above template to obtain the third prompt information. Input the third prompt information into the third generation model to obtain the modified first dialogue information. Thus, rewrite and optimize the original first dialogue information, especially making the first dialogue information more in line with the game environment.

[0099] Continue to refer to Figure 2 , in step S220, obtain the second dialogue information input by the first user or the first user's virtual character in the dialogue group; the first user's virtual character is a virtual character controlled by the first user; the dialogue group includes the above-mentioned multiple non-user virtual characters.

[0100] Among them, the dialogue group is a group chat-enabled group composed of at least three members. In the embodiments of the present disclosure, the members in the dialogue group include the above-mentioned multiple non-user virtual characters, and at least one user or user virtual character. In scenarios such as games and virtual social platforms, users can directly interact with non-user virtual characters in their own identities. For example, users can use their account ID (Identifier, identification) as the speaker to talk to non-user virtual characters, or users can interact with non-user virtual characters in the identity of a virtual character. For example, users use the virtual character they control as the speaker to talk to non-user virtual characters. Therefore, the dialogue group can include users or user virtual characters.

[0101] In one implementation, a conversation group can be established for different users respectively. Exemplarily, in response to the first user or the first user virtual character satisfying a trigger condition, a conversation group including the above-mentioned multiple non-user virtual characters, the first user or the first user virtual character is established, and this conversation group may not include other users or other user virtual characters. For example, in response to the first user inputting second conversation information in the conversation interface of the multiple non-user virtual characters, a conversation group is established, and the first conversation information and the second conversation information are used as the conversation information in this conversation group. Or, in response to the first user virtual character entering a specific area (such as the area where the above-mentioned multiple non-user virtual characters are located), a conversation group is established, the first conversation information is presented in the conversation group, and the first user virtual character can input second conversation information in the conversation group.

[0102] In one implementation, the conversation group can include any number of users or user virtual characters. Exemplarily, a conversation group including the above-mentioned multiple non-user virtual characters can be established. In response to any user or user virtual character satisfying a trigger condition (such as entering the area where the above-mentioned multiple non-user virtual characters are located), this user or user virtual character is added to this conversation group.

[0103] The first user or the first user virtual character can be any user or user virtual character in the conversation group. The second conversation information can be conversation information input in any form such as text, voice, etc.

[0104] Continue to refer to Figure 2 , in step S230, identify the conversation object of the second conversation information.

[0105] Among them, the conversation object of the second conversation information refers to the member in the conversation group targeted by the second conversation information. Exemplarily, the semantics of the second conversation information can be identified, and the conversation object can be determined according to the semantics. Or, a conversation object recognition model can be set, which can be any type of machine learning model such as a neural network. This model can output a multi-classification result representing the conversation object recognition result. The second conversation information is input into the conversation object recognition model, and the conversation object is determined according to the result it outputs. For example, determine the dimension with the highest probability in the multi-classification result output by the conversation object recognition model. If this probability reaches a preset value (this preset value can be determined according to experience or specific circumstances), then determine the object corresponding to this dimension as the conversation object.

[0106] In one implementation, the above-mentioned identifying the conversation object of the second conversation information may include the following steps:

[0107] If the second conversation information includes the name of the target non-user virtual character among the above-mentioned multiple non-user virtual characters, then determine the conversation object of the second conversation information as the target non-user virtual character.

[0108] For example, if the second dialogue information includes the name of NPCa, the dialogue object of the second dialogue information is determined to be NPCa. In this way, the dialogue object can be determined very quickly.

[0109] In one embodiment, if the second dialogue information includes the name of the second user or the name of the second user's virtual character, the dialogue object of the second dialogue information is determined to be the second user or the second user's virtual character. The second user is a user other than the first user in the dialogue group, and the second user's virtual character is a virtual character controlled by the second user. In the case where the dialogue object of the second dialogue information is determined to be the second user or the second user's virtual character, it means that the first user is not having a dialogue with the non-user virtual character, and the subsequent step S240 may not be performed, that is, there is no need to generate the third dialogue information with the non-user virtual character as the speaker.

[0110] In one embodiment, if the second conversation information does not include the name of a non-user virtual character, or does not include the name of a non-user virtual character and the name of a second user or a second user virtual character, it can be determined that the conversation object of the second conversation information is a non-specific object, indicating that the second conversation information is not conversation information for a specific object.

[0111] In one implementation, after obtaining the second dialogue information, the second dialogue information may be preprocessed, and then the third dialogue information may be generated based on the preprocessed second dialogue information. The preprocessing includes but is not limited to one or more of the following methods:

[0112] Foreign character replacement. For example, in a Chinese dialogue scenario, ASCII codes and other methods can be used to detect foreign characters, including English, Japanese, Korean, Russian characters, etc. The detected foreign characters can be translated and replaced with translated Chinese characters, or the detected foreign characters can be replaced with preset characters. Preset characters can be such as "那什麼", "某某", "xx", etc., indicating that the meaning of the character is unclear, which reduces the workload of translation and can increase divergence in the subsequent generation of third dialogue information.

[0113] Legality detection. It refers to detecting whether the second dialogue information contains illegal information, such as sensitive content, non-compliant information, etc. A large amount of legal data and illegal data (which may include open source data and business field data) can be collected and a data set can be constructed. The legality detection model can be trained through the data set. For example, the BERT pre-trained model can be fine-tuned through the data set to obtain a model for text legality classification. The model is used to perform legality detection on the second dialogue information. If it is detected that the second dialogue information includes illegal information, the illegal information can be deleted or blocked, or the subsequent steps can be not performed for the illegal second dialogue information, but fixed information can be returned, such as a non-user virtual character in the dialogue group replying "Sorry, we don't quite understand what you mean" or "Please don't involve sensitive information".

[0114] Continue to refer Figure 2 In step S240, the third dialogue information in the dialogue group is generated according to the second dialogue information and the dialogue object; the speaker of the third dialogue information is a non-user virtual character.

[0115] Among them, the third dialogue information may be a dialogue information in the dialogue group for replying to the second dialogue information. The speaker of the third dialogue information may include the dialogue object of the second dialogue information. Exemplarily, if the dialogue object is a target non-user virtual character, the speaker of the third dialogue information is the target non-user virtual character. If the dialogue object is a non-specific object, the speaker of the third dialogue information may be any one or more non-user virtual characters, such as randomly determining one of the above-mentioned multiple non-user virtual characters as the speaker of the third dialogue information. The reply content may be generated according to the second dialogue information, such as searching the second dialogue information or its keywords in the database of the current scene (the current scene refers to scenes such as games, virtual social platforms, etc., and the database may be a database containing relevant information of the scene) or on the Internet to obtain the reply content. The reply content is combined with the speaker of the third dialogue information to form the third dialogue information. For example, the third dialogue information may be in the form of "speaker: reply content".

[0116] In one embodiment, reference Figure 4 As shown, the above-mentioned generation of the third conversation information in the conversation group according to the second conversation information and the conversation object may include the following steps S410 and S420:

[0117] Step S410, determining fourth prompt information according to the second dialogue information and the dialogue object;

[0118] Step S420: input the fourth prompt information into the fourth generation model, and generate third dialogue information through the fourth generation model.

[0119] Among them, the fourth prompt information is the prompt information used to generate the third dialogue information, and the fourth generation model is the generation model used to generate the third dialogue information. Exemplarily, a fourth prompt template can be obtained, and the second dialogue information, the dialogue object, and the fourth prompt template can be combined to obtain the fourth prompt information.

[0120] In one implementation, the non-user virtual character can be a non-player character in a game. The dialogue processing method may further include the following steps:

[0121] Search for target knowledge information that matches the second dialogue information in the game knowledge information.

[0122] Correspondingly, the above-mentioned determination of the fourth prompt information based on the second dialogue information and the dialogue object may include the following steps:

[0123] Determine the fourth prompt information based on the second dialogue information, the dialogue object, and the target knowledge information.

[0124] Among them, the game-related knowledge information can be used to establish a knowledge base, and the target knowledge information that matches the second dialogue information can be retrieved in the knowledge base. For example, the BGE (a text embedding model) model can be used to encode the second dialogue information and the game knowledge information into vectors, calculate the similarity (such as cosine similarity) between the vector of the second dialogue information and the vector of the game knowledge information, and select one or more game knowledge information with the highest similarity and a similarity reaching the first similarity threshold (a threshold determined in advance, which can be determined according to experience or specific circumstances) as the target knowledge information. Add the target knowledge information to the fourth prompt information to provide a knowledge reference for the game for the fourth generation model. This helps the fourth generation model generate the third dialogue information that conforms to the game scenario and enhances the user's immersion and game experience.

[0125] In one implementation, the dialogue processing method may further include the following steps:

[0126] Search for target memory information that matches the second dialogue information in the memory information of the dialogue group.

[0127] Correspondingly, the above-mentioned determination of the fourth prompt information based on the second dialogue information and the dialogue object may include the following steps:

[0128] Determine the fourth prompt information based on the second dialogue information, the dialogue object, and the target memory information.

[0129] Among them, the memory information of the conversation group is the memory information determined according to the historical conversation information that has occurred in the conversation group. Exemplarily, the conversation information in the conversation group can be segmented and summarized to obtain the memory information corresponding to each segment of conversation information, and added to the memory information of the conversation group. The memory information of the conversation group can reflect the conversation history of the conversation group. The target memory information matching the second conversation information can be searched from it. For example, the BGE model can be used to encode the second conversation information and the memory information of the conversation group into vectors, calculate the similarity (such as cosine similarity) between the vector of the second conversation information and the vector of the memory information, and select one or more memory information with the highest similarity and the similarity reaching the second similarity threshold (a threshold determined in advance, which can be determined according to experience or specific circumstances) as the target memory information. Adding the target memory information to the fourth prompt information can provide a conversation context reference for the process of the fourth generation model to generate the third conversation information. This helps to generate the third conversation information relevant to the context conversation and enhance the coherence and interactivity of the user conversation experience.

[0130] In one implementation, the fourth prompt information can be determined according to the second conversation information, the conversation object, the target knowledge information, and the target memory information.

[0131] Exemplarily, the fourth prompt template is as follows:

[0132] "I want you to act as multiple talk show comedians

[0133] The information of actor 1 is:

[0134] {NPC1 information}

[0135] The information of actor 2 is:

[0136] {NPC2 information} ...

[0138] The information of actor n is:

[0139] {NPCn information}

[0140] Reference knowledge:

[0141] {target knowledge information}

[0142] Chat memory:

[0143] {target memory information}

[0144] Task:

[0145] You will use your wisdom, creativity and observation skills. Please generate a conversation that meets the requirements between the actor and the user based on the above content.

[0146] You should ensure to incorporate personal anecdotes or experiences into daily activities to make them more relevant and appealing to the audience.

[0147] {Second dialogue information}

[0148] {Reply prefix}”

[0149] Among them, more historical dialogue information, such as the last 10 rounds of dialogue information, can be input in {Second dialogue information}, and relevant information about the dialogue object, such as if the dialogue object is a target non-user virtual character, the reply prefix can be "{Target non-user virtual character} says to {First user}:", if the dialogue object is a non-specific object, the program can randomly determine a non-user virtual character as the speaker of the third dialogue information, and the reply prefix can be "{Randomly determined non-user virtual character} says to {Everyone}:". In addition, the target knowledge information and target memory information can be filled into the corresponding positions in the above template, and the background information of non-user virtual characters (i.e., NPC1 information, NPC2 information, etc.) can also be filled in to obtain the fourth prompt information. The fourth prompt information is input into the fourth generation model to obtain the third dialogue information.

[0150] In one implementation, the dialogue processing method may further include the following steps:

[0151] Determine the dialogue state of the second dialogue information;

[0152] Add corresponding preset information to the third dialogue information according to the dialogue state of the second dialogue information.

[0153] Among them, the dialogue state may refer to the state of the speaker of the second dialogue information, such as the emotional state, etc. Exemplarily, if the first user or the first user virtual character inputs the second dialogue information for the first time (i.e., inputs dialogue information in the dialogue group for the first time), the dialogue state of the second dialogue information can be determined as "first entry". If it is not the first time to input the second dialogue information, a sentiment analysis model (such as fine-tuning the BERT pre-trained model through text emotion classification data to obtain a sentiment analysis model) can be used to process the second dialogue information, and positive, negative, neutral and other emotion classification results can be output, that is, the speaking state of the second dialogue information.

[0154] Preset information corresponding to different dialogue states can be set in advance, and the preset information is used to respond to the dialogue state of the second dialogue information. Exemplarily, the preset information can be added to the reply prefix of the third dialogue information. If the dialogue state of the second dialogue information is "first entry", the corresponding preset information can be "{Speaker of the third dialogue information} gives a warm welcome to {First user} and says:"; if the dialogue state of the second dialogue information is positive, the corresponding preset information can be "{Speaker of the third dialogue information} expresses {happy} to {First user / everyone} and says:"; if the dialogue state of the second dialogue information is negative, the corresponding preset information can be "{Speaker of the third dialogue information} expresses {sad} to {First user / everyone} and says:"; if the dialogue state of the second dialogue information is neutral, the corresponding preset information can be empty, that is, the reply prefix is not modified.

[0155] In one implementation, the dialogue state and / or preset information of the second dialogue information can be added to the fourth prompt information. For example, the preset information is added to "{Reply prefix}" in the above fourth prompt template. In this way, the fourth generation model can learn the state of the speaker of the second dialogue information from the dialogue state and preset information, and generate the third dialogue information that matches this state.

[0156] In one implementation, the dialogue processing method may further include the following steps:

[0157] Find a target expression semantically similar to the third dialogue information according to the third dialogue information, and add the target expression to the third dialogue information.

[0158] Exemplarily, a large number of expressions (such as emoji) and their corresponding semantics can be collected to establish an expression semantics database. Match the semantics of the third dialogue information with the semantics of each expression. For example, encode the semantics of the third dialogue information and the semantics of the expression into vectors, and calculate the similarity of the vectors. Select one or more expressions with the highest semantic similarity to the third dialogue information and the similarity reaching the third similarity threshold (a threshold determined in advance, which can be determined according to experience or specific circumstances, such as 0.7) as the target expression. Add the target expression to the third dialogue information, for example, it can be concatenated after the reply content of the third dialogue information. Increase the diversity and liveliness of the third dialogue information.

[0159] In one implementation, post-processing can be performed on the third dialogue information. The post-processing includes but is not limited to one or more of the following methods:

[0160] Foreign character replacement. For example, in a Chinese conversation scenario, if it is detected that the third conversation information generated by the fourth generation model contains foreign characters, the foreign characters can be translated and replaced with the translated Chinese characters, or the detected foreign characters can be replaced with preset characters. The preset characters can be, for example, "that thing", "so-and-so", "xx", etc.

[0161] Legality detection. It refers to detecting whether the third conversation information contains illegal information, such as sensitive content, non-compliant information, etc. A large number of legal data and illegal data (which can include open-source data and business domain data) can be collected and a dataset can be constructed. By using the dataset to train a legality detection model, for example, the BERT pre-trained model can be fine-tuned through the dataset to obtain a model for text legality classification. This model is used to perform legality detection on the second conversation information. If it is detected that the third conversation information includes illegal information, the illegal information can be deleted or blocked, or the third conversation information can be replaced with other reply content, such as "Sorry, we don't quite understand what you mean", etc.

[0162] It should be noted that for a second conversation information (such as a sentence input by the first user), the program can generate any number of third conversation information. For example, it can generate one third conversation information replied by a non-user virtual character, or it can generate multiple third conversation information replied by multiple non-user virtual characters. For example, the first user inputs a sentence in the conversation group, and the program generates three third conversation information, which are replied by NPCa, NPCb, and NPCc respectively, or the program generates one third conversation information replied by NPCa and one third conversation information replied by NPCb, and these two third conversation information are continuous in content.

[0163] In the embodiments of the present disclosure, the second conversation information is input by the first user or the first user virtual character, and the program generates the third conversation information replied by the non-user virtual character. Based on such a form, multiple rounds of conversations can be generated.

[0164] In one embodiment, the conversation processing method may further include the following steps:

[0165] Segment and summarize the conversation information in the conversation group to obtain the memory information corresponding to each segment of the conversation information, and add it to the memory information of the conversation group.

[0166] Among them, the conversation information in the conversation group, including the second conversation information, can be segmented according to the number of conversation information or the time interval, etc. For example, every 10 rounds of conversation information (one round of conversation information can include a second conversation information and the corresponding third conversation information) can be used as one segment of conversation information. Or, if the time interval between two conversation information exceeds a preset interval (which can be determined according to experience or specific circumstances, such as 5 minutes), a segmentation point is set between the two conversation information.

[0167] For each segment of conversation information, key content can be extracted to generate memory information. For example, keywords can be extracted from each segment of conversation information, and memory information can be generated based on the keywords. For example, the memory information can be a combination or encoding of the keywords.

[0168] In one implementation, a corresponding fifth prompt information can be generated according to each segment of conversation information. For example, each segment of conversation information is combined with a fifth prompt template to obtain the fifth prompt information; the fifth prompt information is input into a fifth generation model to obtain the corresponding memory information.

[0169] Exemplarily, the fifth prompt template is as follows:

[0170] “#Task

[0171] You are now a conversation summary expert, and your task is based on the input "historical conversation", and then:

[0172] - Summarize the core content of the current segment of conversation in one sentence as a memory fragment.

[0173] - Analyze the emotional tone of the current conversation and choose one from [pleasant, depressed, average].

[0174] Please ensure that the emotional tone is one of [pleasant, depressed, average].

[0175] #Format

[0176] Summary:

[0177] Emotion: <Emotion>

[0178] #Historical conversation

[0179] {10 rounds of historical conversation fragments}”

[0180] Fill each piece of dialogue information (i.e., a 10-round historical dialogue segment) into the above template to obtain the fifth prompt information. Input the fifth prompt information into the fifth generation model to obtain the memory information corresponding to each piece of dialogue information (i.e., "a summary in one sentence") and be able to determine the emotion of each piece of dialogue information. Record the memory information and emotion in the memory information of the dialogue group for subsequent use. For example, when generating new third dialogue information later, use the recorded memory information.

[0181] The above method of summarizing dialogue information in segments helps the program maintain context consistency and coherence when processing long conversations, while avoiding problems such as omission of key memory information and slow processing due to excessive information processed at one time.

[0182] Figure 5 The schematic flow of the dialogue processing method is shown. Generate the first dialogue information between multiple non-user virtual characters by the generation model. For example, the first prompt information can be generated first, input the first prompt information into the generation model, and output the first dialogue information. The first dialogue information can be displayed in the dialogue group. Obtain the second dialogue information input by the user, perform preprocessing to obtain the preprocessed second dialogue information, and the preprocessed second dialogue information can be displayed in the dialogue group. Obtain the target knowledge information from the knowledge information of the game and the target memory information from the memory information of the dialogue group. Identify the dialogue object of the second dialogue information, determine the fourth prompt information according to the preprocessed second dialogue information, the dialogue object, the target knowledge information, and the target memory information, input the fourth prompt information into the generation model, and output the third dialogue information. Perform postprocessing on the third dialogue information to obtain the postprocessed third dialogue information, which can be displayed in the dialogue group. Summarize the dialogue information in the dialogue group (including the above preprocessed second dialogue information and postprocessed third dialogue information), for example, segment it, generate the fifth prompt information according to each piece of dialogue information, input the fifth prompt information into the generation model, and output the corresponding memory information. Store the memory information generated by the dialogue summary in the memory information of the dialogue group for use when generating subsequent dialogue information.

[0183] An exemplary embodiment of the present disclosure also provides a dialogue processing device. Refer to Figure 6 As shown, the dialogue processing device 600 may include the following program modules:

[0184] The first dialogue generation module 610 is configured to obtain the background information of multiple non-user virtual characters and generate the first dialogue information between the multiple non-user virtual characters according to the background information of the multiple non-user virtual characters;

[0185] A second conversation acquisition module 620, configured to acquire second conversation information input by a first user or a first user virtual character in a conversation group; the first user virtual character is a virtual character controlled by the first user; the conversation group includes the multiple non-user virtual characters;

[0186] A conversation object recognition module 630, configured to recognize a conversation object of the second conversation information;

[0187] A third conversation generation module 640, configured to generate third conversation information in the conversation group according to the second conversation information and the conversation object; the speaker of the third conversation information is the non-user virtual character.

[0188] In one implementation, the generating the first conversation information between the multiple non-user virtual characters according to the background information of the multiple non-user virtual characters includes: determining a conversation topic; determining second prompt information according to the conversation topic and the background information of the multiple non-user virtual characters; inputting the second prompt information into a second generation model, and generating the first conversation information through the second generation model.

[0189] In one implementation, the non-user virtual character is a non-player character in a game; the determining the conversation topic includes: determining first prompt information according to the background information of the game; inputting the first prompt information into a first generation model, and generating the conversation topic through the first generation model.

[0190] In one implementation, the non-user virtual character is a non-player character in a game; the conversation processing device 600 is further configured to: determine third prompt information according to the background information of the game and the first conversation information; input the third prompt information into a third generation model, and generate modified first conversation information through the third generation model.

[0191] In one implementation, the generating the third conversation information in the conversation group according to the second conversation information and the conversation object includes: determining fourth prompt information according to the second conversation information and the conversation object; inputting the fourth prompt information into a fourth generation model, and generating the third conversation information through the fourth generation model.

[0192] In one implementation, the non-user virtual character is a non-player character in a game; the conversation processing device 600 is further configured to: search for target knowledge information matching the second conversation information in game knowledge information; the determining the fourth prompt information according to the second conversation information and the conversation object includes: determining the fourth prompt information according to the second conversation information, the conversation object, and the target knowledge information.

[0193] In one implementation, the dialogue processing device 600 is further configured to: search for target memory information matching the second dialogue information in the memory information of the dialogue group; the determining the fourth prompt information according to the second dialogue information and the dialogue object includes: determining the fourth prompt information according to the second dialogue information, the dialogue object, and the target memory information.

[0194] In one implementation, the dialogue processing device 600 is further configured to: segment and summarize the dialogue information in the dialogue group to obtain memory information corresponding to each segment of dialogue information, and add it to the memory information of the dialogue group.

[0195] In one implementation, the identifying the dialogue object of the second dialogue information includes: if the second dialogue information includes the name of the target non-user virtual character among the multiple non-user virtual characters, determining that the dialogue object of the second dialogue information is the target non-user virtual character.

[0196] In one implementation, the dialogue processing device 600 is further configured to: determine the dialogue state of the second dialogue information; add corresponding preset information to the third dialogue information according to the dialogue state of the second dialogue information.

[0197] In one implementation, the dialogue processing device 600 is further configured to: search for a target expression semantically similar to the third dialogue information according to the third dialogue information, and add the target expression to the third dialogue information.

[0198] The specific details of each part in the above device have been described in detail in the implementation manner of the method part. The details not disclosed can be seen in the implementation manner content of the method part, so they will not be repeated here.

[0199] It should be noted that although several modules or units of a device for action execution are mentioned in the above detailed description, this division is not mandatory. In fact, according to the exemplary implementation manners of the present disclosure, the features and functions of the two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0200] The exemplary implementation manners of the present disclosure further provide a computer program product. The computer program product includes a computer program, and when the computer program is executed by a processor, the above method is implemented.

[0201] In one embodiment, a computer program product may be a tangible product containing a computer program, such as a computer-readable storage medium storing the computer program. The readable storage medium may be a storage medium based on signals such as electricity, magnetism, light, electromagnetic, infrared, etc., including but not limited to: random access memory (RAM), read-only memory (ROM), magnetic tape, floppy disk, flash memory, mechanical hard disk drive (HDD), solid state drive (SSD), and so on. Exemplarily, the computer program product may be implemented as a non-volatile storage medium storing the computer program, such as read-only memory, NAND flash memory, etc.

[0202] In one embodiment, a computer program product may be an intangible product containing a computer program. Exemplarily, the computer program product may be implemented as a virtual digital product, such as an executable file storing the computer program, a digital file such as an installation package.

[0203] The code of the computer program can be written in one or more programming languages. Programming languages such as C, Java, C++, etc. The program code can be executed entirely on the user's computing device, or partially on the user's computing device, or executed as an independent software package, or partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server. In the case involving a remote computing device, the remote computing device can be connected to the user's computing device through any type of network, such as a local area network (LAN), a wide area network (WAN), etc., or can be connected to an external computing device (for example, through an Internet connection provided by an operator).

[0204] The computer program can be carried or transmitted by signals such as electricity, magnetism, light, electromagnetic, infrared, etc. The electronic device can convert the signal carrying the computer program into a digital signal and then run the computer program. When the computer program runs on the electronic device, its code is used to cause the electronic device to execute (more specifically, can cause the processor of the electronic device to execute) the method steps of various exemplary embodiments of the present disclosure. For example, the following steps can be executed: Step S210, obtain background information of multiple non-user virtual characters, and generate first dialogue information between the multiple non-user virtual characters according to the background information of the multiple non-user virtual characters; Step S220, obtain second dialogue information input by the first user or the first user virtual character in the dialogue group; the first user virtual character is a virtual character controlled by the first user; the dialogue group includes the above-mentioned multiple non-user virtual characters; Step S230, identify the dialogue object of the second dialogue information; Step S240, generate third dialogue information in the dialogue group according to the second dialogue information and the dialogue object; the speaker of the third dialogue information is a non-user virtual character.

[0205] Implementing the above method steps through a computer program, generating first conversation information between multiple non-user virtual characters based on the background information of the multiple non-user virtual characters, obtaining second conversation information input by a first user or a first user virtual character in a conversation group, identifying the conversation object of the second conversation information, and generating third conversation information based on the second conversation information and the conversation object. A group chat implementation solution between a user and multiple non-user virtual characters is provided. By introducing the first conversation information and generating third conversation information of the non-user virtual characters for the second conversation information and the conversation object of the second conversation information, the diversity and interest of the conversation information of the non-user virtual characters are improved, the problem of relatively fixed conversation information is solved, and the user experience is enhanced.

[0206] An exemplary embodiment of the present disclosure also provides an electronic device. The electronic device may include a processor and a memory. The memory stores executable instructions of the processor, such as a computer program. The processor executes the method steps of various exemplary embodiments of the present disclosure by executing the executable instructions.

[0207] The following refers to Figure 7 to exemplarily illustrate the electronic device in the form of a general computing device. It should be understood that Figure 7 the electronic device 700 shown is only an example and should not impose limitations on the functions and usage scope of the embodiments of the present disclosure.

[0208] As Figure 7 shown, the electronic device 700 may include: a processor 710, a memory 720, a bus 730, an I / O (input / output) interface 740, and a network adapter 750.

[0209] The memory 720 may include volatile memory, such as RAM 721 and a cache unit 722, and may also include non-volatile memory, such as ROM 723. The memory 720 may further include one or more program modules 724. Such program modules 724 include but are not limited to: an operating system, one or more application programs, other program modules, and program data. The implementation of a network environment may be included in each or some combination of these examples. For example, the program module 724 may include each module in the above device.

[0210] The processor 710 may include one or more processing units. For example, the processor 710 may include processing units such as an AP (Application Processor), a modem processor, a GPU, an ISP (Image Signal Processor), a controller, an encoder, a decoder, a DSP (Digital Signal Processor), a baseband processor, and / or an NPU (Neural-Network Processing Unit).

[0211] The processor 710 can be used to execute executable instructions stored in the memory 720. For example, it can execute the following steps: Step S210, obtain background information of multiple non-user virtual characters, and generate first dialogue information between the multiple non-user virtual characters according to the background information of the multiple non-user virtual characters; Step S220, obtain second dialogue information input by the first user or the first user virtual character in the dialogue group; the first user virtual character is a virtual character controlled by the first user; the dialogue group includes the above-mentioned multiple non-user virtual characters; Step S230, identify the dialogue object of the second dialogue information; Step S240, generate third dialogue information in the dialogue group according to the second dialogue information and the dialogue object; the speaker of the third dialogue information is a non-user virtual character.

[0212] By the processor 710 executing the above method steps, first dialogue information between multiple non-user virtual characters is generated according to the background information of the multiple non-user virtual characters, second dialogue information input by the first user or the first user virtual character in the dialogue group is obtained, the dialogue object of the second dialogue information is identified, and third dialogue information is generated according to the second dialogue information and the dialogue object. An implementation solution for group chat between a user and multiple non-user virtual characters is provided. The first dialogue information is introduced, and third dialogue information of non-user virtual characters is generated for the second dialogue information and the dialogue object of the second dialogue information, improving the diversity and interest of the dialogue information of non-user virtual characters, improving the problem that the dialogue information is relatively fixed, and enhancing the user experience.

[0213] The bus 730 is used to implement connections between different components of the electronic device 700 and may include a data bus, an address bus, and a control bus.

[0214] The electronic device 700 can communicate with one or more external devices 800 (such as a keyboard, a mouse, an external controller, etc.) through the I / O interface 740.

[0215] The electronic device 700 can communicate with one or more networks through the network adapter 750. For example, the network adapter 750 can provide mobile communication solutions such as 3G / 4G / 5G, or wireless communication solutions such as wireless local area network, Bluetooth, and near field communication. The network adapter 750 can communicate with other modules of the electronic device 700 through the bus 730.

[0216] Although Figure 7 not shown in the figure, other hardware and / or software modules can also be provided in the electronic device 700, including but not limited to: a display, microcode, device drivers, redundant processors, external disk drive arrays, tape drives, and data backup storage systems, etc.

[0217] As can be seen from the above, the technical solution of the present disclosure can be implemented as a method, apparatus, system, computer program product, storage medium, electronic device, etc. Those skilled in the art can understand that various aspects of the present disclosure can be specifically implemented in the following forms, namely: a complete hardware implementation, a complete software implementation (including firmware, microcode, etc.), or an implementation combining hardware and software aspects, such as can be respectively referred to as "circuit", "module" or "system".

[0218] It should be understood that the present disclosure is not limited to the specific method steps or structures already described and shown in the drawings, and various modifications and changes can be made without departing from its scope. Based on the specific implementation provided by the present disclosure, those skilled in the art will easily think of other implementation manners. Therefore, the specific implementation provided by the present disclosure is only exemplary, and the scope and spirit of the present disclosure are pointed out by the claims, and should cover any variations, uses, or adaptive changes of the present disclosure, and these variations, uses, or adaptive changes follow the general principles of the present disclosure and include the well-known common knowledge or conventional technical means in the technical field not disclosed by the present disclosure.

Claims

1. A method for processing a conversation, characterized in that: The method comprises: Acquire background information of a plurality of non-user virtual characters, and generate first dialogue information between the plurality of non-user virtual characters according to the background information of the plurality of non-user virtual characters; Acquire second dialogue information input by a first user or a first user virtual character in a dialogue group; the first user virtual character is a virtual character controlled by the first user; the dialogue group includes the plurality of non-user virtual characters; Identifying a conversation partner of the second conversation information; The third dialogue information in the dialogue group is generated according to the second dialogue information and the dialogue object; the speaker of the third dialogue information is the non-user virtual character.

2. The method according to claim 1, characterized in that: The step of generating first dialogue information between the plurality of non-user virtual characters according to the background information of the plurality of non-user virtual characters comprises: Determine the topic of conversation; Determining second prompt information according to the conversation topic and background information of the plurality of non-user virtual characters; The second prompt information is input into a second generation model, and the first dialogue information is generated by the second generation model.

3. The method according to claim 2, characterized in that The non-user virtual character is a non-player character in the game; Determining the topic of conversation includes: Determine first prompt information according to the background information of the game; The first prompt information is input into a first generation model, and the conversation topic is generated by the first generation model.

4. The method according to claim 2, characterized in that: The non-user virtual character is a non-player character in the game; the method further comprises: Determine third prompt information according to the background information of the game and the first dialogue information; The third prompt information is input into a third generation model, and the modified first dialogue information is generated by the third generation model.

5. The method according to claim 1, characterized in that The step of generating the third conversation information in the conversation group according to the second conversation information and the conversation object includes: Determine fourth prompt information according to the second dialogue information and the dialogue object; The fourth prompt information is input into a fourth generation model, and the third dialogue information is generated by the fourth generation model.

6. The method according to claim 5, characterized in that The non-user virtual character is a non-player character in the game; the method further comprises: Searching the game knowledge information for target knowledge information that matches the second dialogue information; The determining the fourth prompt information according to the second dialogue information and the dialogue object includes: The fourth prompt information is determined according to the second dialogue information, the dialogue object, and the target knowledge information.

7. The method according to claim 5, characterized in that The method further comprises: searching the memory information of the conversation group for target memory information matching the second conversation information; The determining the fourth prompt information according to the second dialogue information and the dialogue object includes: The fourth prompt information is determined according to the second dialogue information, the dialogue object, and the target memory information.

8. The method according to claim 7, characterized in that The method further comprises: The conversation information in the conversation group is summarized in sections to obtain memory information corresponding to each section of the conversation information, and the memory information is added to the memory information of the conversation group.

9. The method according to claim 1, characterized in that: The identifying the dialog object of the second dialog information includes: If the second dialogue information includes the name of a target non-user virtual character among the plurality of non-user virtual characters, it is determined that the dialogue object of the second dialogue information is the target non-user virtual character.

10. The method according to claim 1, characterized in that The method further comprises: determining a dialog state of the second dialog information; Corresponding preset information is added to the third dialogue information according to the dialogue state of the second dialogue information.

11. The method according to claim 1, characterized in that: The method further comprises: Search for a target expression with similar semantics to the third dialogue information according to the third dialogue information, and add the target expression to the third dialogue information.

12. A dialogue processing device, characterized in that: The device comprises: A first dialogue generation module is configured to obtain background information of a plurality of non-user virtual characters, and generate first dialogue information between the plurality of non-user virtual characters according to the background information of the plurality of non-user virtual characters; A second dialogue acquisition module is configured to acquire second dialogue information input by a first user or a first user virtual character in a dialogue group; the first user virtual character is a virtual character controlled by the first user; the dialogue group includes the plurality of non-user virtual characters; a dialogue object identification module, configured to identify a dialogue object of the second dialogue information; The third dialogue generation module is configured to generate third dialogue information in the dialogue group according to the second dialogue information and the dialogue object; the speaker of the third dialogue information is the non-user virtual character.

13. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 11 is implemented.

14. An electronic device, characterized in that: include: processor; A memory, configured to store executable instructions of the processor; The processor is configured to perform the method of any one of claims 1 to 11 by executing the executable instructions.

Images