Interactive entertainment dialogue method and device based on dialogue model, medium and product

Abstract

The invention discloses an interactive entertainment dialogue method and device based on a dialogue model, a medium and a product, and relates to the field of data processing. The method comprises the following steps: receiving current dialogue information input by a user in interactive entertainment content; user interaction behavior data and factual data related to the current dialogue information are retrieved from a preset interaction historical record storage module, and a historical record retrieval result is obtained; according to the retrieval result and the timestamp of the current dialogue information, determining a weight coefficient of the retrieval result and the dialogue information; generating weighted data based on the retrieval result, the current dialogue information and the weight coefficient; inputting the weighted data into a preset dialogue model to obtain reply information output by the dialogue model; and sending the reply information to the user. According to the method and the device, the quality and the efficiency of interactive entertainment dialogues can be improved, the problem that the dialogue model enquires unexperienced story plots when the user does not experience all story lines of a certain role is relieved, and better interactive entertainment experience is brought to the user.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to an interactive entertainment dialogue method, device, medium and product based on a dialogue model. Background Art

[0002] With the development of technology, chatbots are widely used in interactive entertainment fields such as games and interactive videos. They are typically based on fine-tuning large language models to achieve personalized dialogue output. The training corpus is divided into factual and conversational. The former provides long-term memory for the model, while the latter shapes the individual style.

[0003] In related technologies, chatbots (conversational models) often experience "hallucinations." This is primarily due to a failure to fully consider the impact of users' actual interaction history in interactive entertainment on character attributes and parameters. For example, in RPG games, users may have varying degrees of favorability towards different characters, and in interactive videos, users experiencing different storylines may lead to different character experiences. This critical information is not effectively incorporated into model training and dialogue generation.

[0004] Lacking this personalized information, the model's generated dialogue is insufficiently grounded, easily creating "illusions" that don't align with reality. For example, in a game, if a user hasn't experienced a character's entire storyline, the dialogue model might mention unexperienced plot points, negatively impacting the user's interactive experience. Summary of the Invention

[0005] In response to the above-mentioned technical problems and defects, the purpose of the present invention is to provide an interactive entertainment dialogue method, device, medium and product based on a dialogue model, which can improve the quality and efficiency of interactive entertainment dialogue, alleviate the problem that users have not experienced the entire story line of a character, but the dialogue model mentions the unexperienced storyline, and bring users a better interactive entertainment experience.

[0006] To achieve the above-mentioned objectives, in a first aspect, the present invention provides an interactive entertainment dialogue method based on a dialogue model, comprising: receiving current dialogue information input by a user in interactive entertainment content, the interactive entertainment content including electronic games or interactive videos; retrieving user interaction behavior data and factual data related to the current dialogue information from a preset interactive history record storage module to obtain historical record retrieval results, the factual data including storylines or character experiences that have occurred in the interactive entertainment content; determining a weight coefficient between the retrieval result and the dialogue information based on the timestamp of the retrieval result and the current dialogue information; generating weighted data based on the retrieval result, the current dialogue information and the weight coefficient; inputting the weighted data into a preset dialogue model to obtain reply information output by the dialogue model; and sending the reply information to the user.

[0007] The present invention employs the above-described method, first receiving the user's current conversation information within interactive entertainment content such as electronic games or interactive videos, ensuring timely response to the user's interaction needs. Next, relevant user interaction behavior data and factual data are retrieved from the interactive history storage module, providing a rich basis for generating reply information, making the reply more personalized and contextually relevant. By determining weight coefficients using timestamps, the importance of information from different time periods is scientifically balanced, allowing both recent and historical information to play an appropriate role in reply generation. Weighted data is generated based on the retrieval results, current conversation information, and weight coefficients, and input into the conversation model. This generates more accurate and relevant reply information, reducing the possibility that the user may not have experienced the entire storyline of a character, while the conversation model mentions unexperienced plotlines. Through the above-described steps, the present invention achieves a smooth interactive experience, enhances the user's sense of participation and immersion in interactive entertainment, improves the quality and efficiency of interactive entertainment conversations, and provides users with a higher-quality interactive entertainment experience.

[0008] In combination with some embodiments of the first aspect, in some embodiments, before receiving the current conversation information input by the user in the interactive entertainment content, it also includes: obtaining the user's operation records and conversation records in the interactive entertainment content; extracting user interaction behavior data based on the operation records, and obtaining factual data based on the conversation records; and storing the user interaction behavior data and factual data as historical information in the form of vector data in the interactive history record storage module.

[0009] Using the technical solution of the above embodiment, before receiving the current conversation information, the user's operation and conversation records in the interactive entertainment are obtained, and the user's interactive behavior data and factual data are extracted and stored in the storage module in vector form. This helps to fully understand the user's behavior and the context of the interactive entertainment content, providing a rich data foundation for generating more accurate responses. By processing and storing different types of data, the user's historical behavior and plot facts can be better integrated, improving the personalization and coherence of the conversation, and enhancing the user's immersion and participation in the interactive entertainment.

[0010] In combination with some embodiments of the first aspect, in some embodiments, user interaction behavior data is stored in an interaction history record storage module in the form of vector data, including: performing word segmentation processing on the user interaction behavior data through a word segmenter to obtain multiple semantic units; mapping the semantic units into low-dimensional vectors through a preset Embedding model; and storing the low-dimensional vectors corresponding to the user interaction behavior data in the interaction history record storage module according to the set index structure and storage rules.

[0011] By implementing the technical solutions of the aforementioned embodiments, user interaction behavior data is stored in the interaction history storage module in a specific manner, effectively converting complex behavior data into low-dimensional vectors for easy storage and retrieval. Through word segmentation, mapping, and appropriate storage, data management becomes more efficient, providing more accurate user behavior information for subsequent conversation generation. This helps improve the conversation model's ability to understand and respond to user behavior, generating responses that better meet user needs and enhancing the quality of the interactive entertainment experience.

[0012] In combination with some embodiments of the first aspect, in some embodiments, the interactive history record storage module includes a vector database, and storing factual data in the interactive history record storage module in the form of vector data includes: utilizing the index building function of the vector database, selecting a vector index algorithm, and creating an index structure based on the characteristics of the factual data; converting the factual data into a low-dimensional vector through a text vectorization tool to obtain factual vector data; storing the factual data in the vector database according to the index structure, and updating the index structure.

[0013] The technical solution of the above embodiment utilizes the capabilities of a vector database to store factual data. An index structure is created based on data characteristics, converting it into low-dimensional vectors and storing them in an ordered manner. This allows for rapid retrieval and utilization of factual data, providing the conversation model with accurate plot context and other information, making responses more logical and relevant. Furthermore, updating the index structure ensures the timeliness and validity of data, improving overall system performance and user experience.

[0014] In combination with some embodiments of the first aspect, in some embodiments, the weight coefficients of the retrieval results and the conversation information are determined based on the timestamps of the retrieval results and the current conversation information, including: sorting the retrieval results and the conversation information in order from the latest timestamp to the earliest one to obtain a historical information sequence; configuring weight coefficients for information elements in the historical information sequence in a manner that gradually decreases the size of the coefficients, the information elements including the retrieval results and the conversation information.

[0015] Adopting the technical solution of the above embodiment, sorting search results and conversation information by timestamp and assigning descending weights can appropriately reflect the varying importance of information at different time stages. This makes recent information more influential in conversation generation while also taking into account the value of historical data, ensuring that responses focus on both the present and historical context, improving the accuracy and rationality of conversations and better adapting to the dynamic changes in interactive entertainment.

[0016] In conjunction with some embodiments of the first aspect, in some embodiments, configuring weight coefficients for information elements in a historical information sequence in a gradually decreasing manner includes: determining the weight coefficients through a weight configuration function, where the weight configuration function includes:

[0017] in, w i represents the weight coefficient, t 0 Represents the timestamp of the current conversation information, t i Represents the timestamp of the information element, a、b、c They are adjustment parameters respectively.

[0018] By adopting the technical solution of the above embodiment, the weight coefficient is determined through a specific weight configuration function, and factors such as timestamp difference and adjustment parameters are comprehensively considered, the value of different information elements can be measured more scientifically. a、b、c The parameter setting makes the determination of weight coefficients more flexible, adapting to the needs of different interactive entertainment scenarios, providing a more reasonable weight distribution for dialogue generation, and improving the quality and pertinence of responses.

[0019] In combination with some embodiments of the first aspect, in some embodiments, weighted data is input into a preset dialogue model to obtain reply information output by the dialogue model, including: determining the user's current emotional state based on the current dialogue information; inputting the current emotional state and weighted data into the dialogue model to obtain reply information.

[0020] By employing the technical solutions of the above embodiments, the user's emotional state is determined based on the current conversation information and input into the conversation model along with weighted data, enabling responses to be more tailored to the user's emotional needs. Taking user emotions into account helps generate more humane and emotionally resonant responses, strengthening the user's emotional connection with interactive entertainment content and increasing user satisfaction and engagement.

[0021] In a second aspect, an embodiment of the present invention provides an electronic device, comprising: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to enable the electronic device to execute the method described in the first aspect or the second aspect, and any possible implementation of the first aspect or the second aspect.

[0022] In a third aspect, the present invention provides a computer-readable storage medium comprising instructions, which, when executed on the electronic device, enables the electronic device to execute the method described in the first aspect or the second aspect, and any possible implementation of the first aspect or the second aspect.

[0023] In a fourth aspect, the present invention provides a computer program product comprising instructions, which, when the computer program product is run on the electronic device, enables the electronic device to execute the method described in the first aspect or the second aspect, and any possible implementation of the first aspect or the second aspect.

[0024] It is understood that the electronic device provided in the second aspect, the storage medium provided in the third aspect, and the computer program product provided in the fourth aspect are all used to execute the method provided by the present invention. Therefore, the beneficial effects that can be achieved can be referred to the beneficial effects of the corresponding methods and will not be repeated here.

[0025] One or more technical solutions provided by the present invention have at least the following technical effects or advantages: 1. By collecting user action and conversation records in interactive entertainment, we extract user interaction behavior data and factual data, storing them in vector form in a storage module. When generating responses, this data is combined with timestamp-based weighting coefficients, allowing the conversation model to fully consider the user's personalized behavior and historical interactions. For example, in a game, if a user frequently selects a specific character for interaction, the system can use this historical behavior to generate responses that are more tailored to the character's characteristics or tasks related to that character when they engage in a conversation again. This allows users to experience a unique and personalized conversational experience in interactive entertainment, enhancing their deep engagement and immersion in the content.

[0026] 2. Use timestamps to determine weight coefficients for search results and conversation information, sorting information chronologically and assigning weights in descending order. This approach scientifically balances the importance of information from different time periods. Recent information tends to better reflect the user's current needs and context, and assigning higher weights makes responses more targeted and timely. Historical information also provides context and coherence for responses, preventing responses from appearing abrupt or inconsistent with the user's historical behavior. For example, in a story-based interactive video, a user's current conversation may be related to previous plot developments. By assigning appropriate weights, the system can generate responses that both align with the current plot and consider the overall storyline, ensuring accuracy and rationality.

[0027] 3. The user's current emotional state is determined based on the current conversation information and input into the conversation model along with weighted data. This allows responses to be tailored not only to the logic and relevance of the content but also to the user's emotional needs. When the user is in a positive mood, the system can generate equally positive and cheerful responses, further enhancing their sense of enjoyment. Conversely, when the user is in a negative mood, comforting and helpful responses are provided, making the user feel cared for by the system. This approach, which considers user emotions, can enhance the emotional resonance between users and interactive entertainment content, improve user satisfaction and engagement, and make interactive entertainment conversations more humane and heartwarming. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The accompanying drawings are incorporated into and constitute a part of the specification, illustrate embodiments consistent with the present invention, and together with the specification, are used to explain the principles of the present invention. Obviously, the drawings described below are only some embodiments of the present invention, and those skilled in the art can derive other drawings based on these drawings without inventive effort. In the drawings: Figure 1 This is a flow chart of an interactive entertainment dialogue method based on a dialogue model according to an embodiment of the present invention; Figure 2 is a flow chart of another interactive entertainment dialogue method based on a dialogue model according to an embodiment of the present invention; Figure 3 It is a schematic diagram of the architecture of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0029] The terms used in the following embodiments of the present invention are for the purpose of describing specific embodiments only and are not intended to limit the present invention. As used in the specification of the present invention, the singular expressions "a," "an," "above," "the," and "this" are intended to include the plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that the term "and / or" as used in the present invention refers to any and all possible combinations of one or more of the listed items.

[0030] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying relative importance or implicitly indicating the quantity of the technical features indicated. Thus, a feature designated "first" or "second" may explicitly or implicitly include one or more of such features. In the description of the embodiments of the present invention, unless otherwise specified, "plurality" means two or more.

[0031] It should also be noted that, unless otherwise clearly specified and limited, in the embodiments of the present invention, terms such as "setting" and "connection" should be understood in a broad sense. For example, "connection" can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, or it can be the internal connection of two components; it can be a wired communication connection or a wireless communication connection. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to the specific circumstances. The embodiments of the present invention are described in detail below.

[0032] An embodiment of the present invention provides an interactive entertainment dialogue method based on a conversation model. This method first comprehensively captures the user's immediate needs by receiving current conversation information from interactive entertainment content such as electronic games or interactive videos. It then accurately retrieves user interaction behavior data and factual data related to the current conversation information from a preset interaction history storage module. This process fully considers the user's actual operational history during the interactive entertainment process, such as the character's action path selected in the game and the plot branches experienced in the interactive video, as well as factual data such as the actual storyline or character experience, thereby avoiding the problem of ignoring this key information in related technologies.

[0033] When determining weight coefficients, the importance of different data types is scientifically measured based on the timestamps of search results and current conversation information, giving more appropriate weight to recent interaction data and conversation information in model decisions. This weighted data is then fed into the pre-set conversation model, ensuring that the model comprehensively considers multiple factors when generating responses. This ensures that responses generated by the conversation model are based on the user's actual operation history and accurate factual data, while also appropriately reflecting the timeliness of different data. This effectively avoids the appearance of content related to scenarios not experienced by the user, thereby resolving the potential "hallucination" issue with chatbots and significantly improving the user's interactive entertainment experience, enhancing the coherence, logic, and immersion of the conversation.

[0034] The following combination Figure 1 The method of this embodiment is specifically described, which can be applied to an interactive entertainment system (hereinafter referred to as the system), and specifically includes the following steps: Step 201: Receive current dialogue information input by a user in interactive entertainment content.

[0035] The interactive entertainment content includes electronic games or interactive videos, and the current conversation information includes text messages or voice messages.

[0036] Specifically, the interactive entertainment system receives text input or voice commands from users in electronic games or interactive videos through integrated input interfaces such as keyboards, mice, touch screens, voice recognition systems or game controllers, converts the voice information into text through voice recognition technology, and further processes the user's current conversation information for subsequent interaction and response.

[0037] Step 202 : Retrieve user interaction behavior data and factual data related to the current conversation information from a preset interaction history record storage module to obtain a history record retrieval result.

[0038] Among them, factual data includes storylines or character experiences that have occurred in interactive entertainment content. It is objective information that has occurred in interactive entertainment content, such as the development of the storyline, character experiences, game world settings, etc. in the game, as well as the established plot and character background presented in the interactive video, etc., which are used to provide accurate factual basis and background knowledge for the dialogue.

[0039] User interactive behavior data refers to the behavioral records generated by user operations in interactive entertainment content (such as electronic games or interactive videos), including but not limited to character selection, task completion status, scene interaction behavior, choice of plot branches, etc., reflecting the user's personalized operation process during the entertainment process.

[0040] In this embodiment, the system first performs a comprehensive semantic analysis of the current conversation, using natural language processing techniques such as lexical analysis, syntactic analysis, and semantic understanding algorithms to extract key elements of the conversation, such as key topics, entities, and events. For example, if a conversation mentions a specific character or scene in a game, the system can accurately identify it.

[0041] Next, based on these extracted key elements, the system conducts a deep search within the user interaction behavior data storage area. By pattern-matching key elements with stored user action records, such as character actions, scene transitions, and task completion status, relevant user interaction behavior data is filtered out. Simultaneously, within the factual data storage area, the system searches for corresponding factual data such as past storylines and character experiences based on the key elements. For example, if the current conversation involves a specific game plot point, the system will quickly locate all relevant factual data for that plot point.

[0042] In order to improve retrieval efficiency and accuracy, the interactive history record storage module adopts advanced indexing technologies such as inverted index and vector index, and builds index structure according to the characteristics and semantics of the data, so that the retrieval process can quickly locate relevant data.

[0043] In addition, the system will sort and filter the search results based on the user's historical usage habits and the context of the current conversation, giving priority to presenting the most relevant and most likely valuable user interaction behavior data and factual data for the current conversation, thereby obtaining accurate historical record retrieval results and providing strong support for subsequent conversation processing.

[0044] In this embodiment, the interaction history storage module is an important component for storing user interaction history records. It mainly includes an interaction history database and a cache. The interaction history database is an organized data set that can store a large amount of interaction history data in a long-term and stable manner, such as user chat records, like records, comment records, etc. on a certain platform. This data is structured and stored in the database for subsequent efficient retrieval, query, and analysis. The cache is a memory that temporarily stores data and improves data access speed. When a user frequently accesses certain interaction history records, this data will be temporarily stored in the cache. The next time the same data is accessed, the system will prioritize reading from the cache without having to query the database again, greatly improving the system's response speed and enhancing the user experience. The cache and database work together to form the interaction history storage module, providing the system with efficient and fast data storage and access services.

[0045] Step 203: Determine the weight coefficient of the retrieval result and the conversation information according to the timestamps of the retrieval result and the current conversation information.

[0046] Specifically, the system first obtains the timestamps of all search results and the current conversation information and converts them into relative time values ​​based on the current time. It then categorizes the time intervals based on their timeliness. For example, you can set several time intervals, such as the last hour, the last day, or the last week. Each interval is assigned a different initial weight. Recent time intervals have a relatively higher weight, while more distant time intervals have a gradually decreasing weight.

[0047] At the same time, adjustments can also be made based on the type and importance of the data. If the search results are factual data related to key plot points or important events, or data from frequent user interactions, their weight can be increased appropriately. Conversely, if the data is relatively common, old, and less important, its weight can be reduced.

[0048] In addition, the system can also be dynamically adjusted based on user behavior patterns. If a user frequently interacts with a specific type of data within a certain time period, the weight of that type of data within the corresponding time range can be appropriately increased.

[0049] Through this step, the system can more reasonably determine the weight coefficients of retrieval results and conversation information, thereby better integrating data from different time points in the subsequent conversation generation process and providing users with more accurate and personalized conversation responses.

[0050] Step 204: generating weighted data based on the search results, current conversation information, and weight coefficients.

[0051] First, for each piece of data in the search results, whether it is user interaction data or factual data, it is adjusted according to its corresponding weight coefficient. If the weight coefficient is higher, the data is more important in the weighting process.

[0052] Next, the current conversation information is treated as an independent data item and assigned a corresponding weight (which can be determined based on factors such as its relevance to the search results and its recency). Next, for numerical data, the data value can be directly multiplied by its weight coefficient to obtain the weighted value. For textual data, word embedding can be used. After converting the text into a word embedding representation, the value of each dimension of each word embedding is multiplied by the corresponding weight coefficient to obtain the weighted word embedding.

[0053] Finally, all weighted search result data and the current conversation information data are integrated, either by summing or averaging, to obtain the final weighted data. This weighted data combines data from different sources and adjusts it appropriately based on its importance, providing more valuable information for subsequent input into the conversation model.

[0054] Step 205: input the weighted data into a preset dialogue model to obtain reply information output by the dialogue model.

[0055] First, the conversational model analyzes the weighted data, identifying different types of data sources and characteristics. If the weighted data includes user interaction behavior data, the conversational model analyzes user behavior patterns and preferences in interactive entertainment to infer potential user intentions and needs. For factual data, the model extracts key information, such as storyline and character experience, as the basis for responses and background knowledge.

[0056] The conversation model then uses its internal neural network architecture to perform complex calculations and transformations on the weighted data. Through layer-by-layer neural network transmission and activation functions, the input data is integrated and transformed into latent semantic representations. Based on these semantic representations, the model then generates preliminary response candidates based on its pre-learned language patterns and conversation strategies. There may be multiple candidate responses, and the model further evaluates their plausibility, coherence, and relevance to the user input.

[0057] Finally, the most appropriate response is selected as the output. This response takes into account the user's current conversation information and combines the weighted influence of the retrieved historical data. It can better meet the user's conversation needs in interactive entertainment and provide users with a more natural and personalized conversation experience.

[0058] Among them, the dialogue model is built based on technologies such as natural language processing and deep learning. It can receive weighted data including user interaction behavior data, current dialogue information and corresponding weight coefficients as input. Relying on its internal complex neural network architecture formed by a large amount of corpus training, it first deeply analyzes these input data to explore the key elements such as user behavior preferences and storylines contained therein, and then uses the multi-layer neuron calculation and activation mechanism to fuse and semantically transform the data. Then, based on the mastered language rules, dialogue logic and response strategies for interactive entertainment scenarios learned during the training process, it generates multiple candidate responses. Finally, through rigorous evaluation of the performance of each candidate response in terms of rationality, coherence and relevance to user input, it selects and outputs the response information that best meets user needs, can be integrated into the interactive entertainment background and can provide users with a natural and personalized dialogue experience.

[0059] Step 206: Send the reply information to the user.

[0060] Specifically, once the conversation model generates a reply, the system sends it to the user via the appropriate output channel. If the user enters the conversation message in text form, the system presents the reply in clear, neat text format within a specific dialogue display box within the game or interactive video interface, ensuring that the text is well-formatted and easy to read, allowing the user to intuitively see the reply. If the user uses voice input, the device uses speech synthesis technology to convert the text into smooth, natural speech based on the reply's text content using a pre-trained speech synthesis model, selecting appropriate parameters such as timbre and intonation. This is then played through the device's speakers, allowing the user to conveniently hear the reply. For devices that support multimodal interaction, the system may also incorporate elements such as images and animations to present the reply in a more vivid and visual manner within the corresponding interactive interface. For example, in-game characters may make expressions and movements that match the reply content, comprehensively conveying the reply to the user and ensuring that the user receives it promptly and clearly, and is immersed in the interactive entertainment conversation.

[0061] The method of this embodiment utilizes the aforementioned steps. First, by receiving the user's current conversation information within interactive entertainment content and retrieving relevant user interaction behavior data and factual data from a pre-set storage module, it fully considers the user's actual operational history within the game or interactive video, avoiding the chatbot's "illusion" of a plot not experienced by the user, significantly improving the user experience. Secondly, by determining the weight coefficients between the retrieval results and the conversation information based on timestamps, it scientifically and rationally balances the importance of recent interactions and historical data, making the output of the conversation model more targeted and accurate. The response information generated based on this weighted data integrates the user's personalized operational history with their current conversation needs, providing users with a more natural and personalized conversation service. Furthermore, the response information is sent to users in a variety of formats, adapting to different interaction scenarios and enhancing the user's interactivity and immersion in the interactive entertainment content. Overall, this invention significantly improves and innovates the conversation experience in the interactive entertainment field.

[0062] The following combination Figure 2 To further illustrate the method of this embodiment, it specifically includes the following steps: Step 301: Obtain the user's operation records and conversation records in the interactive entertainment content.

[0063] Specifically, during the interactive entertainment process, whether the user manipulates the character in the electronic game to move, fight, complete tasks, and other actions, or makes plot branch choices and character interaction operations in the interactive video, all will be monitored and recorded by the system in real time, which constitute the operation record. At the same time, the text, voice, and other forms of conversations between the user and the characters and systems in the game or video will also be fully saved, which is the conversation record. By embedding a data acquisition module in the corresponding interactive entertainment program, the system accurately captures the instruction information corresponding to each operation action and the specific content of each conversation, ensuring comprehensive and accurate acquisition of the user's operation records and conversation records in the entire interactive entertainment content, laying the foundation for the subsequent extraction of valuable data.

[0064] Step 302: extract user interaction behavior data based on the operation records, and obtain factual data based on the conversation records.

[0065] Specifically, collected operation records (which exist in the form of text, log files, or structured data in a specific format) are cleaned to remove redundant information, erroneous records, and incomplete data fragments to ensure data quality and integrity. For example, abnormal command records caused by network fluctuations are removed from game operation records. Next, data parsing technology is used to break down the operation records into meaningful basic operation units based on the specific protocols and data formats followed by different interactive entertainment content. For example, in a role-playing game, character movement operations are parsed into specific parameters such as direction and speed, and skill release operations are parsed into elements such as skill name, release timing, and target object.

[0066] Then, with the help of feature extraction algorithms, such as rule-based feature extraction (extracting corresponding features based on pre-set rule templates related to user behavior, such as "the first time a character uses a newly unlocked skill after each upgrade") and machine learning-based feature extraction methods (using unsupervised learning algorithms such as cluster analysis to cluster similar operation behaviors into one category to explore potential behavioral characteristics), key behavioral feature information is extracted from these basic operation units.

[0067] Finally, the extracted characteristic information reflecting user interactive behavior, such as character growth trajectory, task completion order, skill usage preferences, and scene interaction characteristics, is integrated and standardized, organized according to a unified data structure to form complete user interactive behavior data for subsequent storage, analysis, and utilization.

[0068] In some embodiments, embedding (a technology that maps high-dimensional data to low-dimensional space) technology can be used to store user interaction behavior data in the form of vector data into the interaction history storage module, specifically including: S3021, performing word segmentation processing on the user interaction behavior data through a word segmenter to obtain multiple semantic units.

[0069] A tokenizer is used to analyze the collected and organized user interaction data (which may be recorded in text format, such as character actions and quest completion status). Based on specific language rules and a predefined vocabulary, the tokenizer breaks down a segment of user interaction data into relatively independent units with clear semantic meaning, known as semantic units. For example, for user interaction data such as "The player completed the exploration of the third chapter of the main quest in the game and frequently used the character's healing skills to help teammates," the tokenizer will segment it into multiple semantic units, such as "player," "game," "completed," "main quest," "chapter three," "exploration," "frequently," "use," "character," "healing skills," "help," and "teammates." This prepares the way for the subsequent mapping of these semantic units into vectors, making the data more easily processed and analyzed.

[0070] S3022: Map the semantic unit into a low-dimensional vector through a preset Embedding model.

[0071] Among them, Embedding models are usually trained with a large amount of corpus data, such as common models such as Word2Vec and BERT, which have the ability to convert semantic units into low-dimensional vector representations.

[0072] Each previously obtained semantic unit is sequentially input into the embedding model. Based on its internally learned semantic relationships and vector representation rules, the model calculates and generates a corresponding low-dimensional vector for each semantic unit. For example, after processing the semantic unit "healing skills," it is mapped into a low-dimensional vector with a specific numerical combination. In this vector space, the distances and angles between the vectors corresponding to different semantic units can, to a certain extent, reflect their semantic similarity and relevance. This allows the once abstract and complex semantic units to be represented in a quantifiable, computer-processable vector form, facilitating subsequent data storage and utilization.

[0073] S3023: Store the low-dimensional vector corresponding to the user interaction behavior data into the interaction history storage module according to the set index structure and storage rules.

[0074] First of all, the index structure is often constructed based on factors such as the characteristics, sources or application scenarios of the data. For example, indexes are established according to the chronological order of user operations and the behaviors corresponding to different game modules. The purpose is to facilitate the subsequent quick and accurate retrieval of the required vector data. The storage rules clarify the storage format, storage location and related metadata management methods of the data. During specific storage, the low-dimensional vectors corresponding to the user interaction behavior data will be placed in the corresponding position in the interaction history storage module according to the index structure, and the related metadata will be recorded at the same time, such as the user ID corresponding to the vector data, operation time and other key information, to ensure that the data can be stored in order in the storage module and easy to find, thereby providing strong data support for subsequent retrieval, analysis and dialogue generation based on these vector data.

[0075] On the other hand, the process of obtaining factual data based on conversation records is as follows: first, the conversation records are preprocessed, and conversation records in different formats (text, text after speech-to-text conversion, etc.) are uniformly converted into a standard text format, and text normalization operations are performed, such as removing unnecessary punctuation, correcting typos, and unifying capitalization, to ensure the standardization and accuracy of the text content.

[0076] Subsequently, the lexical analysis technology in natural language processing (NLP) is used to segment the text, split the sentences into independent words, and perform part-of-speech tagging to determine the part-of-speech category of each word, such as noun, verb, etc., which helps with subsequent semantic understanding.

[0077] Next, syntactic analysis techniques are used to construct a grammatical structure for the sentence, clearly presenting the syntactic relationships between words and aiding in understanding the sentence's semantics. Based on this, semantic understanding and entity recognition technologies are employed. Using pre-trained semantic models (such as language models based on the deep learning Transformer architecture, which have been trained on large amounts of corpus and possess strong semantic understanding capabilities), entities mentioned in the text (such as in-game character names and place names) and the relationships between them are identified. This allows for the discovery of factual information such as past events in the game, the characters' past experiences, and the setting of the world they inhabit.

[0078] Finally, the various types of extracted factual information are integrated and deduplicated, and stored according to a certain knowledge graph or structured data organization form to ensure the systematicness and integrity of the factual data, so that it can be used as reliable background knowledge in subsequent dialogue generation and other links.

[0079] In some embodiments, the interaction history storage module specifically includes a vector database. RAG (Retrieval-Augmented Generation, a deep learning model architecture that combines retrieval and generation) technology can be used to store factual data in the interaction history storage module in the form of vector data, including: S3024, using the index building function of the vector database, selecting a vector indexing algorithm, and creating an index structure based on the characteristics of the factual data.

[0080] The first step is to create an index structure. Vector databases have specialized indexing capabilities, which help us manage and retrieve data more efficiently. We need to choose an appropriate vector indexing algorithm based on the characteristics of the factual data. Common examples include tree-based indexing algorithms (such as KD trees and ball trees), hash-based indexing algorithms, and graph-based indexing algorithms. If the factual data is relatively low-dimensional and well-distributed, a tree-based indexing algorithm may be more convenient for data segmentation and location. However, if the data volume is large and fast search efficiency is more important, a hash-based indexing algorithm may be more suitable. Based on the selected algorithm, the index structure is created by combining characteristics of the factual data, such as plot elements, character backgrounds, and game world settings. This allows subsequent queries for vector data related to a specific plot or for vectors corresponding to a character background to quickly locate the corresponding data, laying the foundation for efficient data storage and retrieval.

[0081] S3025, converting the factual data into a low-dimensional vector using a text vectorization tool to obtain factual vector data.

[0082] Text vectorization tools are generally built based on deep learning or machine learning principles. Commonly used tools include Word2Vec, FastText, and BERT. For example, when factual data is fed into a selected text vectorization tool, the tool analyzes and processes each word, phrase, and even entire sentence in the text based on the semantic and grammatical rules learned through pre-training, as well as vector representation methods, mapping them into low-dimensional vectors with specific numerical values. For example, for a factual data description like "In the game, the protagonist obtains a mysterious magic item in a mysterious valley," a text vectorization tool will output a corresponding low-dimensional vector. The position of this vector in vector space and its relationship to other vectors can, to a certain extent, reflect its semantic information. This transforms complex factual data, originally in text form, into factual vector data that is easily stored, calculated, and compared by computers.

[0083] S3026: Store the factual data into the vector database according to the index structure, and update the index structure.

[0084] The vector database provides a dedicated storage space and management mechanism for storing these vector data. According to the storage location, organization method and other requirements specified by the index structure, each factual vector data is placed in the corresponding position to ensure the orderliness of data storage. At the same time, as new data is continuously stored, in order to ensure that the index structure can always accurately and efficiently serve subsequent data retrieval work, it needs to be updated. For example, when new plot-related factual vector data is added, the part of the index structure related to the plot index may need to adjust the node relationship or add new index branches, etc., so that it can promptly reflect the latest status of the data in the database. In this way, when the vector corresponding to a certain factual data needs to be queried later, the updated index structure can quickly locate the corresponding data, ensuring the efficient operation of the entire storage and retrieval system, so that the interactive history storage module can better utilize these vector data to provide support for application scenarios such as interactive entertainment dialogue.

[0085] Step 303: The user interaction behavior data and factual data are taken as historical information and stored in the interaction history record storage module in the form of vector data.

[0086] Specifically, after obtaining user interaction behavior data and factual data, corresponding data vectorization technology will be adopted to facilitate the subsequent efficient storage, retrieval, and utilization of this data. For example, using commonly used models or algorithms such as Word2Vec and BERT, each specific behavioral feature in the user interaction behavior data and each semantic unit in the factual data are mapped into a low-dimensional vector representation, so that the originally complex and diverse text or behavior information can exist in a form that is easy to calculate and compare in the vector space. Afterwards, these converted vector data are stored in the corresponding storage areas of the interactive history record storage module according to established storage rules. The interactive history record storage module can manage these vector data in an orderly manner to ensure that relevant data can be retrieved quickly and accurately when needed later, thereby providing strong data support for the entire interactive entertainment dialogue method.

[0087] Step 304: Receive current dialogue information input by the user in the interactive entertainment content.

[0088] In order to receive the current dialogue information input by users in interactive entertainment content, the system is equipped with a variety of input interfaces and processing mechanisms adapted to different interaction scenarios.

[0089] When a user is in the virtual world of a video game or immersed in an interactive video, if they enter text via the keyboard, the system monitors keyboard signals in real time, accurately capturing and integrating the user's keystrokes into complete conversational sentences, ensuring accurate character recognition and proper order, for clear and complete information about the current conversation. If the user communicates via voice, a highly sensitive microphone continuously captures sound signals. The voice recognition module then rapidly processes these signals, using advanced speech recognition algorithms to convert the speech into text. It also analyzes voice characteristics such as intonation and speed to assist in understanding the user's emotions and intent. On touch-enabled systems, if the user taps on preset dialogue options or input boxes, the system uses touch sensing to accurately capture the user's touch location and gestures, converting them into corresponding conversational information. Regardless of the user's input method, the system can promptly and accurately receive and interpret the current conversational information entered within the interactive entertainment content, providing a reliable foundation for subsequent conversation processing.

[0090] In this step, if the conversation is naturally interrupted while the user is inputting the conversation information, the following methods can be used to handle it: S3041, by monitoring changes in input device signals and combining timers and other means, real-time monitoring and identification of natural interruptions that occur when users input conversation information.

[0091] Specifically, first, the system needs to have the ability to monitor the user's input status in real time. It determines whether an input interruption has occurred by monitoring signal changes from input devices (such as keyboards and microphones). For text input, if it is detected that no new characters are input for a period of time during the normal input process, it may indicate a natural interruption. For voice input, if the voice signal suddenly stops and no further sound is heard within a set short waiting time, it can also be considered a natural interruption. This process can be assisted by technical means such as timers. By setting a reasonable time threshold, once no input action is made after the threshold, a natural interruption is confirmed.

[0092] S3042, when a natural interruption of the conversation is recognized, the conversation information entered by the user is immediately and completely saved.

[0093] Specifically, for text input, all characters currently entered into the input box are recorded completely to ensure that nothing is missed. For voice input, speech recognition technology is used to instantly save the text content that has been collected and converted up to the moment of interruption. Even if the original intended words are not fully expressed later, this part of the valid information is retained so that subsequent processing can be based on it.

[0094] S3043, by combining factors such as the user's past conversation habits, operation behaviors and the current interactive entertainment scene context, analyze the reasons for conversation interruption to infer user intentions.

[0095] Specifically, based on users' past conversation habits, the system will establish a user conversation history database, recording information such as conversation content, topic types, and speaking frequency in different interactive entertainment scenarios. Using data mining techniques, users' conversation patterns are extracted, such as commonly used vocabulary, sentence structure, and topic preferences. When a natural interruption occurs, the system compares the current input with historical conversation patterns. If it finds that the current input topic or language style is similar to the user's previous performance in similar situations, for example, when discussing a specific game level, the user often paused to think before continuing, and the current conversation is also interrupted during a discussion of a similar level, it may be inferred that the user is making similar adjustments in thinking.

[0096] The system continuously monitors and records every user action within interactive entertainment, including task completion, skill usage, and scene interaction. Using behavioral analysis algorithms, it identifies user patterns and habits. For example, if a user frequently engages in conversational sharing after completing a complex task, an interruption might indicate that the user is preparing to express their opinion about the task but is still formulating their words. Furthermore, through machine learning, the system continuously learns and updates its understanding of the relationship between user actions and conversational intent.

[0097] Integrating the current interactive entertainment scene context, the system analyzes various elements in the game or interactive video in real time, such as ongoing plot events, task status, prompt information, and so on. Natural language processing technology is used to parse the text information in the scene and extract key information. When an interruption occurs, the system correlates the current input with the scene context for analysis. For example, if the user input is interrupted at a key turning point in the plot, it may be that they want to express their views on the plot development but have not yet determined how to express them. In addition, through multimodal information fusion technology, the user's operation behavior, conversation history, and scene context information are integrated to comprehensively determine the cause of the interruption and infer the user's intention, so as to provide users with more accurate guidance and feedback, ensuring that the conversation process can continue smoothly.

[0098] S3044: Based on the inference of the interruption reason and the user's intention, provide the user with corresponding guidance and feedback to assist the user in continuing to enter complete and valid dialogue information.

[0099] First, the system uses natural language generation technology to generate targeted prompts based on the inferred user intent. For example, if it infers that the user interrupted input because they were confused by a new game element, the system can generate prompts such as "Are you confused about the [name of the game element] that just appeared? Please tell me your question." to guide the user to continue speaking.

[0100] For voice input scenarios, the system can use speech synthesis technology to convert prompts into natural and fluent voices and play them to the user, making it easier for the user to continue inputting after hearing the prompt. At the same time, in terms of interface design, the system can highlight relevant information or options based on user intent, assisting the user with visual guidance. For example, if it is speculated that the user may want to discuss a specific game task, relevant images, brief descriptions, and other information about the task can be displayed around the input interface to stimulate further thinking and input from the user.

[0101] Furthermore, the system leverages machine learning algorithms to adjust guidance and feedback based on the user's historical input patterns and preferences. If a user has responded well to a certain type of prompt in the past, the system will prioritize that type in similar situations. By continuously monitoring user responses and adjusting strategies, the system can provide more personalized and effective guidance and feedback, helping users continue to enter complete and valid conversational information, ensuring a smooth flow of conversation in interactive entertainment.

[0102] Step 305 : Retrieve user interaction behavior data and factual data related to the current conversation information from a preset interaction history record storage module to obtain a history record retrieval result.

[0103] This step can refer to the description of the above embodiment and will not be repeated here.

[0104] Step 306: Sort the search results and conversation information in order of timestamps from the latest to the earliest, to obtain a historical information sequence.

[0105] Among them, each data item in the search results and conversation information will be marked with a corresponding timestamp, which accurately records the time node when it was generated or appeared.

[0106] The system extracts these timestamps and performs sorting based on a specialized sorting algorithm. This algorithm treats all search results and conversation messages as elements to be sorted, arranging them in order from the latest timestamp, based on the chronological order represented by the timestamps. For example, more recent conversation messages or recently retrieved data, due to their later timestamps, will be placed at the front of the sequence, while older elements will be arranged in descending order. By sequentially comparing the timestamps of each element and adjusting their positions, the search results and conversation messages are ultimately integrated into a complete historical information sequence, providing an orderly data foundation for subsequent operations such as configuring weight coefficients and generating conversations based on this data.

[0107] Step 307 : configuring weight coefficients for information elements in the historical information sequence in a gradually decreasing manner, where the information elements include search results and conversation information.

[0108] The system first sets an initial weight coefficient, typically assigned to the first (most recent) information element in the historical message sequence. Subsequently, the weight coefficients of subsequent information elements are gradually reduced according to a predetermined decreasing rule. This decreasing rule can be linear, such as setting a fixed difference and subtracting the weight coefficient for each subsequent information element. Alternatively, it can be exponential or other functions that follow a trend of decreasing coefficients. For example, the most recent conversation message at the beginning is assigned a weight coefficient of 0.8, while subsequent information elements receive progressively smaller weight coefficients, such as 0.7, 0.6, and so on, following the decreasing rule. Regardless of the specific decreasing method used, the principle is that newer information elements (i.e., those closer to the current moment) should have a greater impact on the current conversation, so they are assigned relatively higher weight coefficients. As information ages, its weight coefficients gradually decrease. This effectively reflects the varying importance of information from different time periods in the conversation generation process, thereby ensuring that subsequent responses generated based on these weighted information are more realistic and reasonable.

[0109] In some embodiments, the weight coefficient is determined specifically by a weight configuration function, the weight configuration function including:

[0110] in, w i represents the weight coefficient, t 0 Represents the timestamp of the current conversation information, t i Represents the timestamp of the information element, a、b、cThey are adjustment parameters respectively.

[0111] In the formula, Simulates the time decay effect, that is, as the timestamp of the information element t i Timestamp of the current conversation message t 0 increases, the weight coefficient w i This is consistent with human perception of the importance of information, that is, the closer the information is to the current time, the more relevant and important it is.

[0112] In the function a、b、c These are adjustable parameters that allow the weight calculation to be customized according to specific application scenarios and needs. α Control the maximum value of the weight to ensure that the weight coefficient does not exceed a certain upper limit. β Adjusts the rate of time decay. This can be a positive or negative number. A positive number means the weight decreases over time, while a negative number means the weight increases over time. c This ensures that even for very old information elements, their weight coefficients will not drop to zero. This provides a basic weight for all information elements, ensuring their basic influence in the dialogue model.

[0113] Step 308: generating weighted data based on the search results, current conversation information, and weight coefficients.

[0114] This step can refer to the description of the above embodiment and will not be repeated here.

[0115] Step 309: Determine the user's current emotional state based on the current conversation information.

[0116] When a user enters a conversation, the system first applies sentiment analysis techniques from natural language processing (NLP) to determine their emotional state. The system then preprocesses the conversation, performing tasks such as text normalization, removing unnecessary punctuation, and standardizing capitalization to ensure the text format is convenient for subsequent analysis. Next, lexical analysis is used to segment the text into individual words, and part-of-speech tagging is performed to clearly identify each word's part of speech.

[0117] Next, syntactic analysis is used to construct a grammatical structure for the sentence, assisting in understanding the overall semantics of the sentence. Based on this, sentiment classification models based on machine learning or deep learning are used to determine emotional state. These models may be trained on large amounts of text data with labeled sentiment (e.g., positive, negative, neutral, etc.). For example, sentiment classification models built using support vector machines (SVMs), deep learning convolutional neural networks (CNNs), and long short-term memory networks (LSTMs) are used. The models extract emotional keywords, modal particles, and semantic features of the sentences in the conversation as a basis for judgment. For example, if positive words such as "happy" and "joyful" appear and the overall semantics convey positive meaning, the sentence is classified as positive. If negative words such as "sad" and "angry" are included and the sentence semantics match, the sentence is classified as negative. If there are no words or semantics with obvious sentiment, the sentence is classified as neutral. This accurately determines the user's current emotional state based on the current conversation information.

[0118] Step 310: Input the current emotional state and weighted data into the dialogue model to obtain reply information.

[0119] The conversation model itself utilizes a neural network architecture. After receiving these two inputs, it first analyzes and extracts features. For the current emotional state, the model identifies the specific emotion category (positive, negative, or neutral) and its intensity, adjusting the emotional tone of the response. For example, for users with positive emotions, the model tends to generate equally positive and cheerful responses; for users with negative emotions, the model considers providing comforting and helpful responses.

[0120] The user interaction behavior data, factual data, and their corresponding weight coefficients contained in the weighted data will be used by the model to grasp the user's actual operations in interactive entertainment, familiarize themselves with existing storylines and other background knowledge, and ensure that the response fits the user's behavior trajectory and the plot background.

[0121] Then, the model relies on the calculations and activations between its numerous internal neurons to deeply integrate and transform the input information based on pre-learned language patterns, dialogue strategies, and response rules for different emotional scenarios, generating multiple candidate responses.

[0122] Finally, by evaluating the coherence, rationality, and relevance of each candidate response to the user's input, the most appropriate response is selected as the final output. This response information takes into account the user's current emotional feelings and combines past relevant data, thereby providing users with more personalized and tailored conversation content.

[0123] Step 311: Send the reply information to the user.

[0124] This step can refer to the description of the above embodiment and will not be repeated here.

[0125] The method provided by an embodiment of the present invention utilizes the aforementioned steps to first receive a user's current conversation information within interactive entertainment content, such as electronic games or interactive videos, ensuring timely response to the user's interaction needs. Next, relevant user interaction behavior data and factual data are retrieved from an interactive history storage module, providing a rich basis for generating responses, making responses more personalized and contextually relevant. By determining weight coefficients based on timestamps, the importance of information from different time periods is scientifically balanced, allowing both recent and historical information to play an appropriate role in response generation. Weighted data is generated based on the retrieval results, current conversation information, and weight coefficients, and input into a conversation model. This generates responses that are more accurate and more relevant to the user's actual situation, reducing the possibility that a chatbot or conversation model mentions unexperienced storylines while the user has not experienced the entire storyline of a character. Through the aforementioned steps, the present invention achieves a smooth interactive experience, enhances user engagement and immersion in interactive entertainment, improves the quality and efficiency of interactive entertainment conversations, and provides users with a higher-quality interactive entertainment experience.

[0126] In some embodiments, the interactive entertainment system includes an electronic device. The following describes the electronic device in the embodiment of the present invention from the perspective of hardware processing. Figure 3 , which is a schematic diagram of a physical device structure of an electronic device in an embodiment of the present invention.

[0127] It should be noted that Figure 3 The structure of the electronic device shown is only an example and should not limit the functions and scope of use of the embodiments of the present invention.

[0128] like Figure 3 As shown, the electronic device includes a central processing unit (CPU) 401, which can perform various appropriate actions and processes, such as the methods described in the above embodiments, based on programs stored in a read-only memory (ROM) 402 or programs loaded from a storage unit 408 into a random access memory (RAM) 403. RAM 403 also stores various programs and data required for system operation. CPU 401, ROM 402, and RAM 403 are interconnected via a bus 404. An input / output (I / O) interface 405 is also connected to bus 404.

[0129] The following components are connected to the I / O interface 405: an input section 406 including an audio input device, push button switches, and the like; an output section 407 including a liquid crystal display (LCD), an audio output device, indicator lights, and the like; a storage section 408 including a hard disk and the like; and a communication section 409 including a network interface card such as a LAN (Local Area Network) card or a modem. The communication section 409 performs communication processing via a network such as the Internet. A drive 410 is also connected to the I / O interface 405 as needed. Removable media 411, such as a magnetic disk, an optical disk, a magneto-optical disk, or a semiconductor memory, is installed in the drive 410 as needed, so that computer programs read from the removable media can be installed in the storage section 408 as needed.

[0130] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program including a computer program for executing the methods illustrated in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 409 and / or installed from removable media 411. When executed by central processing unit (CPU) 401, the computer program performs the various functions defined in the present invention.

[0131] It should be noted that specific examples of computer-readable storage media may include, but are not limited to, an electrical connection having one or more conductors, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0132] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present invention. Each box in the flowchart or block diagram can represent a module, program segment, or part of the code, and the above-mentioned module, program segment, or part of the code contains one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings.

[0133] Specifically, the electronic device of this embodiment includes a processor and a memory, the memory is coupled to one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and one or more processors call the computer instructions to enable the electronic device to execute the method provided by the above embodiment.

[0134] As another aspect, the present invention further provides a computer-readable storage medium, which may be included in the electronic device described in the above embodiments, or may exist independently and not incorporated into the electronic device. The storage medium carries one or more computer programs, and when executed by a processor of the electronic device, the electronic device implements the methods provided in the above embodiments.

[0135] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments can still be modified, or some of the technical features thereof can be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present invention.

[0136] As used in the above embodiments, the term “when” may be interpreted to mean “if” or “after” or “in response to determining that” or “in response to detecting that”, depending on the context. Similarly, the phrases “upon determining that” or “if (stated condition or event) is detected” may be interpreted to mean “if determining that” or “in response to determining that” or “upon detecting (stated condition or event)” or “in response to detecting (stated condition or event)”, depending on the context.

[0137] Those skilled in the art will appreciate that all or part of the process steps in the above-described method embodiments can be implemented by a computer program instructing the relevant hardware. The program can be stored in a computer-readable storage medium, and when executed, the program can include the process steps in the above-described method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as ROM or random access memory (RAM), magnetic disks, or optical disks.

Claims

1. An interactive entertainment dialogue method based on a dialogue model, characterized in that: include: receiving current conversation information input by a user in interactive entertainment content, wherein the interactive entertainment content includes electronic games or interactive videos; Retrieving user interaction behavior data and factual data related to the current conversation information from a preset interaction history storage module to obtain a history retrieval result, wherein the factual data includes storylines or character experiences that have occurred in the interactive entertainment content; Determining a weight coefficient between the search result and the conversation information according to the timestamps of the search result and the current conversation information; generating weighted data based on the search result, the current conversation information, and the weight coefficient; Inputting the weighted data into a preset dialogue model to obtain reply information output by the dialogue model; The reply information is sent to the user.

2. The method according to claim 1, characterized in that Before receiving the current dialogue information input by the user in the interactive entertainment content, the method further includes: Obtaining the user's operation records and conversation records in the interactive entertainment content; Extracting the user interaction behavior data based on the operation record, and acquiring the factual data based on the conversation record; The user interaction behavior data and the factual data are taken as historical information and stored in the interaction history record storage module in the form of vector data.

3. The method according to claim 2, characterized in that Storing the user interaction behavior data in the form of vector data into the interaction history storage module includes: Performing word segmentation processing on the user interaction behavior data through a word segmenter to obtain multiple semantic units; Mapping the semantic unit into a low-dimensional vector through a preset Embedding model; The low-dimensional vector corresponding to the user interaction behavior data is stored in the interaction history storage module according to a set index structure and storage rules.

4. The method according to claim 2, characterized in that The interaction history record storage module includes a vector database, and storing the factual data in the interaction history record storage module in the form of vector data includes: Using the index building function of the vector database, selecting a vector indexing algorithm, and creating an index structure according to the characteristics of the factual data; The factual data is converted into a low-dimensional vector using a text vectorization tool to obtain factual vector data; The factual data is stored in the vector database according to the index structure, and the index structure is updated.

5. The method according to any one of claims 1 to 4, characterized in that The determining, based on the timestamps of the retrieval result and the current conversation information, a weight coefficient of the retrieval result and the conversation information includes: Sorting the search results and the conversation information in descending order of the timestamps to obtain a historical information sequence; The weight coefficients are configured for the information elements in the historical information sequence in a manner of gradually decreasing coefficient sizes, wherein the information elements include the search results and the dialogue information.

6. The method according to claim 5, characterized in that The configuring the weight coefficients for the information elements in the historical information sequence in a manner of gradually decreasing coefficient sizes specifically includes: The weight coefficient is determined by a weight configuration function, wherein the weight configuration function includes: ； in, w i represents the weight coefficient, t 0 Represents the timestamp of the current conversation information, t i a timestamp representing the information element, α, β, γ They are adjustment parameters respectively.

7. The method according to claim 1, characterized in that Inputting the weighted data into a preset dialogue model to obtain reply information output by the dialogue model includes: Determining the current emotional state of the user based on the current conversation information; The current emotional state and the weighted data are input into the dialogue model to obtain the reply information.

8. An electronic device, characterized in that: including one or more processors and memory; The memory is coupled to the one or more processors, and is configured to store computer program codes, where the computer program codes include computer instructions. The one or more processors call the computer instructions to enable the electronic device to execute the method according to any one of claims 1 to 7.

9. A computer-readable storage medium storing computer instructions, characterized in that: When the computer instructions are executed on an electronic device, the electronic device is caused to execute the method according to any one of claims 1 to 7.

10. A computer program product, characterized in that When the computer program product is run on an electronic device, the electronic device is enabled to perform the method according to any one of claims 1 to 7.

Images