Radio content generation method and device based on user portrait and emotion perception, equipment and medium

Abstract

This application discloses a method, apparatus, device, and medium for generating radio content based on user profiles and emotion perception, relating to the field of media content recommendation technology. The method involves: acquiring user data containing profiles, interactions, real-time environment, and current query information, then inputting this data into a preset radio content generation model for processing to generate personalized radio content. This fully integrates long-term user preferences, immediate emotions, and the environment, solving the problem of insufficient personalization in traditional radio content, improving the relevance of content to user needs, and optimizing the radio service experience.

Description

Technical Field

[0001] This invention relates to the field of media content recommendation technology, and in particular to a method, apparatus, device, and medium for generating radio content based on user profiles and emotion perception. Background Technology

[0002] Existing intelligent radio systems primarily rely on large language models to generate content. These systems analyze user queries to generate corresponding audio content, such as podcasts and music recommendations. While these systems can meet users' basic needs to a certain extent, they fail to fully consider users' long-term preferences and immediate emotions when generating content. The systems typically generate content based only on simple user queries, lacking a deep understanding of users' true needs and emotional states, and failing to adequately consider the impact of external environmental factors on user emotions and needs. For example, users' emotions and needs may vary at different times or under different weather conditions, but existing systems cannot adjust content generation and recommendation strategies according to these contextual changes. Therefore, there is an urgent need for a radio content generation method to improve the personalization and contextual relevance of content generation. Summary of the Invention

[0003] The main purpose of this application is to provide a method, apparatus, device, and medium for generating radio content based on user profiles and emotion perception, aiming to solve the technical problem of how to generate radio content that better meets the personalized needs of users through user profiles and emotion perception.

[0004] To achieve the above objectives, this application proposes a radio content generation method based on user profiles and emotion perception, comprising:

[0005] Acquire user data, wherein the user data includes profile data, interaction data, real-time environmental data, and current query information;

[0006] The user data is input into a preset radio content generation model for processing to obtain personalized radio content. The preset radio content generation model includes a user profile construction module, an emotion perception and fusion module, a radio content recommendation module, and a radio content generation module.

[0007] The step of inputting the user data into a preset radio content generation model for processing to obtain personalized radio content includes:

[0008] The user profile data is used to construct a textual user profile through the user profile construction module, and the current query information is parsed to obtain the user's request text.

[0009] The user's emotion tag is obtained by performing emotion recognition on the interaction data through the emotion perception and fusion module and then fusing and reconstructing it with the real-time environmental data.

[0010] The radio content recommendation module sorts the user request text, the user emotion tags, and the textual user profile to obtain recommended radio content.

[0011] The user command is obtained by integrating the textualized user profile, the user emotion tag, the current query information, and the recommended radio content.

[0012] The user's instructions are input into the radio content generation module to generate personalized radio content.

[0013] In one embodiment, the step of obtaining user data includes:

[0014] Send a data retrieval command to the user profile database so that the user profile database returns profile data, wherein the profile data includes user identifier, age, gender, interests and preferences and personality type;

[0015] Collect interaction data generated during the interaction between the user and the radio station, including text data input by the user and voice interaction data;

[0016] Send an environmental data acquisition instruction to a third-party environmental service platform so that the third-party environmental service platform can return real-time weather data and real-time time period data of the user's current location, wherein the real-time environmental data includes the real-time weather data and the real-time time period data;

[0017] Receive the query content input by the user through the radio interactive interface and determine the query content as the current query information;

[0018] The user profile data, the interaction data, the real-time environment data, and the current query information are format-validated and integrated to form user data.

[0019] In one embodiment, the step of constructing a textualized user profile from the profile data using the user profile construction module and parsing the current query information to obtain the user's request text includes:

[0020] The profile data is input into the user profile building module for processing to obtain key fields in the profile data, wherein the key fields include age, gender, interests and preferences, and personality type;

[0021] The key fields are filled into the preset profile generation template to generate the initial user profile text, wherein the profile generation template is constructed by the user profile construction module.

[0022] The initial user profile text is semantically optimized to obtain a textualized user profile;

[0023] The current query information is input into the user profile building module for parsing to obtain the core intent and key entities in the current query information, wherein the core intent includes play, search, and query.

[0024] The core intent and key entities are organized according to a preset structured format to obtain the user requirement text.

[0025] In one embodiment, the step of performing emotion recognition on the interaction data through an emotion perception and fusion module and then fusing and reconstructing it with the real-time environmental data to obtain a user emotion tag includes:

[0026] The voice interaction data in the interaction data is input into the voice emotion recognition unit in the emotion perception and fusion module for feature extraction and emotion classification, and the first emotion probability vector is output.

[0027] The text data in the interaction data is input into the text sentiment classification unit in the emotion perception and fusion module for semantic analysis and sentiment judgment, and a second emotion probability vector is output.

[0028] According to a preset priority rule, a target emotion probability vector is selected from the first emotion probability vector and the second emotion probability vector as the user emotion vector;

[0029] Based on a preset environmental emotion mapping rule base, the real-time weather data and real-time time period data in the real-time environmental data are matched to obtain weather emotion vectors and time emotion vectors respectively.

[0030] The environmental emotion vector is obtained by calculating the weather emotion vector and the time emotion vector.

[0031] The user emotion vector and the environment vector are weighted and fused to obtain a fused emotion vector;

[0032] The fused emotion vectors are sorted by probability, and the fused emotion vector with the highest probability is selected as the user's emotion label.

[0033] In one embodiment, the step of sorting the user request text, the user emotion tag, and the textualized user profile through the radio content recommendation module to obtain recommended radio content includes:

[0034] The user request text and the user emotion tag are input into the coarse ranking unit in the radio content recommendation module to perform semantic matching on all content items in the preset radio content library, and obtain the matching degree between each content item and the user request text and the user emotion tag;

[0035] Select content items with a matching degree higher than a preset matching threshold to form a coarse candidate list;

[0036] The textual user profile and the coarse-ranked candidate list are input into the fine-ranking unit of the radio content recommendation module for calculation, to obtain the profile matching score and intent matching score of each content item in the coarse-ranked candidate list with interest and preference features, wherein the interest and preference features are extracted from the textual user profile by the fine-ranking unit;

[0037] Based on the weighted average of the profile matching score and intent matching score, the candidates are comprehensively sorted in descending order to obtain the refined ranking result;

[0038] Select a preset number of content items from the ranking results to determine the recommended radio station content.

[0039] In one embodiment, the step of integrating the textualized user profile, the user sentiment tag, the current query information, and the recommended radio content to obtain the user instruction includes:

[0040] Determine the integration order of the textualized user profile, the user sentiment tags, the current query information, and the recommended radio content;

[0041] The current query information, the textualized user profile, the user emotion tags, and the recommended radio content are sequentially concatenated according to the integration order to form the initial instruction text;

[0042] The duplicate information in the initial instruction text is deduplicated to obtain the first instruction text;

[0043] Based on a preset instruction format template, an instruction identifier and a timestamp are added to the first instruction text to obtain the second instruction text;

[0044] The second instruction text is subjected to integrity verification. When the second instruction text meets the preset requirements, the second instruction text is determined to be a user instruction.

[0045] In one embodiment, the step of inputting the user instruction into the radio content generation module to generate personalized radio content includes:

[0046] The user command is input into the radio content generation module for semantic parsing to obtain core requirement information;

[0047] The core requirement information is matched with a preset content generation rule base to determine the corresponding content generation strategy, wherein the content generation strategy includes tone and content structure.

[0048] According to the content generation strategy, the user instruction is sent as an input prompt to the large language model to obtain the initial radio content text;

[0049] The initial radio content text is formatted and optimized to determine personalized radio content.

[0050] Furthermore, to achieve the above objectives, this application also proposes a radio content generation device based on user profiles and emotion perception, wherein the radio content generation device based on user profiles and emotion perception includes:

[0051] The acquisition module is used to acquire user data, which includes user profile data, interaction data, real-time environment data, and current query information;

[0052] The result module is used to process the user data by inputting it into a preset radio content generation model to obtain personalized radio content. The preset radio content generation model includes a user profile construction module, an emotion perception and fusion module, a radio content recommendation module, and a radio content generation module. It is also used to construct a textualized user profile from the profile data through the user profile construction module and parse the current query information to obtain the user's request text; to perform emotion recognition on the interaction data through the emotion perception and fusion module and fuse it with the real-time environmental data to obtain user emotion tags; to sort the user's request text, user emotion tags, and textualized user profile through the radio content recommendation module to obtain recommended radio content; to integrate the textualized user profile, user emotion tags, current query information, and recommended radio content to obtain user instructions; and to input the user instructions into the radio content generation module to generate personalized radio content.

[0053] In addition, to achieve the above objectives, this application also proposes a storage medium, which is a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the radio content generation method based on user profiles and emotion perception as described above.

[0054] In addition, to achieve the above objectives, this application also provides a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the radio content generation method based on user profiles and emotion perception as described above.

[0055] This application obtains user data containing profiles, interactions, real-time environment, and current query information, then processes it by inputting it into a preset radio content generation model to generate personalized radio content. This fully combines users' long-term preferences, immediate emotions, and environment, solving the problem of insufficient personalization in traditional radio content, improving the relevance of content to user needs, and optimizing the radio service experience. Attached Figure Description

[0056] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0057] Figure 1 This is a flowchart illustrating the first embodiment of the radio content generation method based on user profiles and emotion perception in this application;

[0058] Figure 2 This is a block diagram of the preset radio content generation model structure of the first embodiment of the radio content generation method based on user profile and emotion perception in this application;

[0059] Figure 3 This is a flowchart illustrating the second embodiment of the radio content generation method based on user profiles and emotion perception in this application;

[0060] Figure 4 This is an environmental emotion mapping rule diagram of the second embodiment of the radio content generation method based on user profiles and emotion perception in this application;

[0061] Figure 5 This is a schematic diagram of the module structure of the radio content generation device based on user profiling and emotion perception in this application.

[0062] Figure 6 This is a schematic diagram of the device structure of the hardware operating environment involved in the radio content generation method based on user profiles and emotion perception in the embodiments of this application.

[0063] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0064] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0065] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0066] Existing intelligent radio systems primarily rely on large language models to generate content. These systems analyze user queries to generate corresponding audio content, such as podcasts and music recommendations. While these systems can meet users' basic needs to a certain extent, they fail to fully consider users' long-term preferences and immediate emotions when generating content. The systems typically generate content based only on simple user queries, lacking a deep understanding of users' true needs and emotional states, and failing to adequately consider the impact of external environmental factors (such as time and weather) on user emotions and needs. For example, users' emotions and needs may differ at different times or under different weather conditions, but existing systems cannot adjust their content generation and recommendation strategies accordingly.

[0067] Therefore, this application proposes a radio content generation method based on user profiles and emotion perception to solve the above problems. The main solution of this application's embodiments is: acquiring user data, wherein the user data includes profile data, interaction data, real-time environmental data, and current query information;

[0068] The user data is input into a preset radio content generation model for processing to obtain personalized radio content. The preset radio content generation model includes a user profile construction module, an emotion perception and fusion module, a radio content recommendation module, and a radio content generation module. The step of inputting the user data into the preset radio content generation model for processing to obtain personalized radio content includes:

[0069] The user profile data is used to construct a textual user profile through the user profile construction module, and the current query information is parsed to obtain the user's request text. The interaction data is then fused and reconstructed using the emotion perception and fusion module to obtain user emotion tags. The user request text, user emotion tags, and textual user profile are sorted by the radio content recommendation module to obtain recommended radio content. The textual user profile, user emotion tags, current query information, and recommended radio content are then integrated to obtain user instructions. The user instructions are then input into the radio content generation module to generate personalized radio content.

[0070] Based on the above, this application also provides a method for generating radio content based on user profiles and emotion perception, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the radio content generation method based on user profiling and emotion perception according to this application. In this embodiment, the radio content generation method based on user profiling and emotion perception includes steps S10 to S20:

[0071] Step S10: Obtain user data.

[0072] It should be noted that user data includes profile data, interaction data, real-time environmental data, and current query information. The specific acquisition steps include: First, sending a data retrieval command to the user profile database, causing the database to return profile data, which includes user identifier, age, gender, interests, and personality type. Profile data is the core of personalized user information, containing key information such as user identifier, age, gender, interests, and personality type. The user identifier is used to uniquely identify the user, ensuring data accuracy and the realization of personalized services. Age and gender are basic user attributes that influence content preferences to some extent. Interest preferences more specifically reflect users' liking for different types of content, such as "photography," "hiking," and "independent films," directly reflecting users' hobbies and interests in life. The system can provide users with content recommendations on related topics based on these interests, enabling users to quickly find radio content that interests them. At the same time, user-defined preferences are also very important, including "science fiction," "folk music," "documentary podcasts," and "classical music." This preference data reflects users' specific preferences in content consumption, and the system can generate more personalized radio content tailored to their tastes based on these preferences. Personality types further describe the user's psychological characteristics and behavioral tendencies, providing a basis for generating content that matches the user's personality traits.

[0073] Next, the system collects interaction data generated during the user's interaction with the radio station. This data includes both text input and voice interaction data. Text data typically comes from user input in the radio app's chat box, search box, etc., directly reflecting the user's immediate needs and interests. Voice interaction data is generated when the user interacts with the radio station using voice commands. After being converted into text using speech recognition technology, it can also be used to analyze the user's emotions and intentions. The collection of interaction data allows the system to capture the user's dynamic behavior in real time, thereby better understanding the user's current state and needs.

[0074] Next, an environmental data acquisition command is sent to a third-party environmental service platform, prompting the platform to return real-time weather and time-period data for the user's current location. This real-time environmental data includes both weather and time-period data, acquired by sending requests to the platform. Real-time weather data indicates the current weather conditions at the user's location, such as sunny, cloudy, or rainy. This information influences the user's mood and content preferences. For example, in rainy weather, users might be more inclined to listen to neutral music or soothing podcasts. Real-time time-period data refers to the specific time of day the user uses the radio, such as early morning, morning, afternoon, evening, or late at night. Different time periods also affect the user's mood and needs; for example, early morning might require energetic content to wake the user, while late at night might be more suitable for quiet, neutral content. By acquiring this environmental data, the system can better adapt to the user's actual situation and generate content that better suits their needs.

[0075] Then, the system receives the query content input by the user through the radio's interactive interface and identifies it as the current query information. Specifically, the user submits a specific query request through the radio's interactive interface (such as voice input, text input, etc.), such as "play a neutral piece of music" or "find a podcast about technology." The system receives these queries in real time and identifies them as the current query information. This process involves natural language processing technology to accurately understand the user's intent and needs and transform them into structured information that can be processed by subsequent modules. The current query information is a direct reflection of the user's immediate needs, providing the system with a clear direction for generation.

[0076] Finally, the user profile data, interaction data, real-time environment data, and current query information are format-validated and integrated to form user data. Specifically, before data integration, the system needs to validate the format of each type of data to ensure its integrity and accuracy. For example, it checks whether the age in the profile data is a valid value and whether the query information conforms to semantic specifications. Data that does not meet the format requirements will be corrected or marked by the system to avoid errors in subsequent processing. The validated data will be integrated into a unified user data structure for subsequent module calls and processing. The integrated user data includes complete user profile information, emotional state, real-time environment information, and current query information requirements. This integration makes the data more organized and improves the overall efficiency of the system.

[0077] Step S20: Input user data into a preset radio content generation model for processing to obtain personalized radio content.

[0078] It should be noted that the user data is input into a preset radio content generation model for processing. This process involves the collaborative work of multiple modules to ensure that the generated radio content meets the user's personalized needs and context. Figure 2 The diagram showing the pre-defined radio content generation model reveals that it comprises a user profile building module, an emotion perception and fusion module, a radio content recommendation module, and a radio content generation module. First, the user profile building module receives profile information from user data and transforms it into natural language descriptive text using specific algorithms and templates. This process involves not only integrating basic user information but also in-depth analysis of user interests, preferences, and personality traits, providing accurate user profiles for subsequent content generation. Next, the emotion perception and fusion module analyzes text and voice information from user interaction data, combined with real-time environmental data (such as weather and time period), to perceive the user's emotional state and generate emotion tags. This module utilizes advanced voice emotion recognition and text emotion analysis technologies to accurately capture the user's immediate emotions, while further optimizing the emotion perception results by incorporating the influence of environmental factors. Subsequently, the radio content recommendation module uses a complex algorithm to filter the most suitable content based on the user profile, emotion tags, and current query information. This module combines the user's long-term preferences and immediate needs, employing a two-stage matching algorithm (coarse intent ranking and fine profile ranking) to extract radio content from a massive content library that matches both the user's current emotions and long-term interests. Finally, the radio content generation module integrates the outputs of the above modules into the final personalized radio content. This module uses natural language generation technology, combined with user profile descriptions, emotion tags, and recommended content, to generate fluent, natural, and emotionally resonant radio content. The generated content not only reflects the user's current needs but also enhances the user's immersion and interactivity through emotional language. This entire process fully utilizes various dimensions of user data, achieving efficient generation of personalized radio content through the synergy between modules, significantly improving the user experience.

[0079] This embodiment acquires user data including profiles, interactions, real-time environment, and current query information, then processes it by inputting it into a preset radio content generation model to generate personalized radio content. This fully combines users' long-term preferences, immediate emotions, and environment, solving the problem of insufficient personalization in traditional radio content, improving the relevance of content to user needs, and optimizing the radio service experience.

[0080] Based on the first embodiment of this application, in the second embodiment of this application, the content that is the same as or similar to that in Embodiment 1 above can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to... Figure 3 The radio content generation method based on user profiles and emotion perception, step S20, further includes steps S201 to S205:

[0081] Step S201: The user profile data is used to construct a textual user profile through the user profile construction module, and the current query information is parsed to obtain the user's demand text.

[0082] It should be noted that this process involves in-depth processing of user profile data and accurate analysis of users' immediate needs.

[0083] Further, step S201 includes: First, inputting the profile data into the user profile building module for processing to obtain key fields from the profile data. Specifically, this process aims to extract key fields from the raw data obtained from the user profile database. These key fields include age, gender, interests and preferences, and personality type. For example, a user might be tagged as 25 years old, male, with a personality type of "INFP," and interests in photography, hiking, and independent films. Through specific processing logic and algorithms, the module can quickly locate and extract this information, which is crucial for subsequent content generation, laying the foundation for subsequent steps.

[0084] Next, the key fields are filled into a preset user profile generation template to generate initial user profile text. This template is specifically designed by the user profile building module, and its structure and content are designed to transform the key fields into concise and descriptive initial user profile text. For example, the template might generate text like: "This user is a 25-year-old male with an INFP personality type who enjoys photography, hiking, and independent films." This template can flexibly adapt to the data characteristics of different users, ensuring that the generated initial text accurately reflects the user's core characteristics and preferences. By filling in the key fields, the generated initial user profile text provides raw material for subsequent semantic optimization.

[0085] Next, the initial user profile text undergoes semantic optimization to obtain a text-based user profile. Specifically, this process involves natural language processing (NLP) technology, using semantic enhancement techniques to polish the initial text, making it more natural and fluent, while strengthening the description of user characteristics. For example, the optimized text will describe the user's interests and preferences in more detail; for instance, it might state, "This 25-year-old male user is mild-mannered, creative, loves photography, enjoys finding inspiration in nature, is passionate about hiking, enjoys outdoor adventures, and also has a strong interest in independent films." Semantic optimization not only improves the text's readability but also enhances its expressive power and guiding role in subsequent content generation. The resulting text-based user profile more accurately reflects the user's personality and needs, providing an important basis for generating personalized radio content.

[0086] Then, the current query information is input into the user profile building module for parsing to obtain the core intent and key entities within the query. Specifically, the current query information directly reflects the user's immediate needs; for example, a user might input "play a relaxing folk song" or "recommend a podcast about technology." Through natural language processing technology, the module can quickly identify the core intent and key entities in the user's query. Core intents include playing, searching, and asking, while key entities are content related to these behaviors, such as song titles, podcast themes, and artist names. For example, for the query "play a relaxing folk song," the core intent is "play," and the key entity is "relaxing folk song." By parsing the query information, the system can accurately understand the user's immediate needs, providing a clear direction for subsequent content recommendation and generation.

[0087] Finally, the core intent and key entities are organized according to a preset structured format to obtain the user request text. Specifically, this structured format aims to transform the user's immediate needs into formatted text that the system can efficiently understand and process. For example, the user request text might be organized as: "Core intent: Play; Key entity: Relaxing folk music." The user request text not only clearly expresses the user's current intent but also, through a structured approach, allows it to be combined with the user profile text, providing accurate input for the final personalized radio content generation. Through the above processing steps, the system can fully utilize user profiles and immediate query information to generate highly personalized radio content that meets user needs, significantly improving the user experience and the system's intelligence level. For example, when a user (25-year-old male, INFP type, enjoys photography, hiking, and independent films) enters the query "Play a relaxing folk song," the system first extracts key fields to construct a textualized user profile: "This 25-year-old male user is mild-mannered, creative, loves photography, enjoys finding inspiration in nature, is passionate about hiking, enjoys outdoor adventures, and also has a strong interest in independent films." Then, the query information is parsed to obtain the core intent "Play" and the key entity "Relaxing folk music." Finally, the system combines user profiles and user needs to recommend a song that aligns with both the user's long-term interests (such as folk music) and their current mood (neutral music). In this way, the system not only meets the user's current needs but also enhances long-term user satisfaction.

[0088] Step S202: The emotion recognition and fusion module performs emotion recognition on the interaction data and combines it with real-time environmental data to reconstruct the user's emotion label.

[0089] It should be noted that the core function of the emotion perception and fusion module is to accurately capture the user's emotional state and adjust and optimize it in combination with real-time environmental factors, thereby generating an emotion label that reflects the user's current emotional state.

[0090] Furthermore, step S202 also includes: First, inputting the voice interaction data from the interaction data into the voice emotion recognition unit in the emotion perception and fusion module for feature extraction and emotion classification, and outputting a first emotion probability vector. Specifically, the voice emotion recognition unit identifies the user's emotional state by analyzing features such as tone, speed, and pitch in the voice signal. For example, if a user inputs "The weather is so nice today, I feel really good," the voice emotion recognition unit will extract features from the voice and classify them into emotion categories such as "happy," "sad," and "angry," generating a 7-dimensional emotion probability vector, such as [0.8, 0.1, 0.0, 0.0, 0.0, 0.0, 0.1], where the probability of "happy" in the first dimension is 0.8.

[0091] Next, the text data from the interaction data is input into the text sentiment classification unit in the emotion perception and fusion module for semantic analysis and sentiment judgment, outputting a second emotion probability vector. Specifically, the text sentiment classification unit identifies the user's emotional state by analyzing the emotional words and semantic tendencies in the text. For example, if a user types "I've been feeling a bit irritable lately, I want to listen to some relaxing music" in the chat box, the text sentiment classification unit will analyze the emotional words and semantic tendencies in the text and generate another 7-dimensional emotion probability vector, such as [0.1, 0.2, 0.0, 0.0, 0.0, 0.0, 0.7], where the probability of "sadness" in the second dimension is 0.2.

[0092] Next, a target emotion probability vector is selected from the first and second emotion probability vectors as the user's emotion vector according to a preset priority rule. Specifically, the priority rule can be set based on the accuracy and reliability of the voice and text data. For example, if the voice data is more accurate, the first emotion probability vector is selected as the user's emotion vector. Assuming that the voice data has a higher priority, the system will select [0.8, 0.1, 0.0, 0.0, 0.0, 0.0, 0.1] as the user's emotion vector.

[0093] Subsequently, based on a pre-defined environmental sentiment mapping rule base, matching is performed on real-time weather data and real-time time period data within the real-time environmental data to obtain weather sentiment vectors and time sentiment vectors, respectively. Specifically, as... Figure 4 The environmental emotion mapping rule diagram shown indicates that the current weather is "sunny" and the time period is "morning". According to the environmental emotion mapping rule base, the emotion vector for sunny weather is [0.5, 0.0, 0.0, 0.1, 0.0, 0.0, 0.4], and the emotion vector for morning is [0.1, 0.0, 0.1, 0.1, 0.0, 0.0, 0.7].

[0094] Then, the weather sentiment vector and the time sentiment vector are calculated to obtain the environmental sentiment vector. The values ​​of the corresponding dimensions of the two vectors are added together and the average value is taken to obtain the environmental sentiment vector that comprehensively reflects the current environmental emotional tendency.

[0095] The user emotion vector and the environment vector are then weighted and fused to obtain a fused emotion vector. Specifically, the cosine similarity between the user emotion vector and the environment emotion vector is first calculated, and the weights are adjusted according to a preset similarity threshold. If the similarity is higher than the threshold, the weight of the user emotion vector is given primary weight; if it is lower than the threshold, the influence of the environment emotion vector is reduced, and then the fused emotion vector is obtained through weighted calculation.

[0096] Finally, the fused emotion vectors are sorted by probability, and the fused emotion vector with the highest probability is selected as the user's emotion label. For example, the dimension of "happiness" has the highest probability after fusion, so "happiness" is determined as the user's emotion label, providing an emotional basis for subsequent radio content recommendation and generation.

[0097] Step S203: The radio content recommendation module sorts the user demand text, user emotion tags, and textual user profile to obtain recommended radio content.

[0098] It should be noted that the radio content recommendation module accurately recommends radio content that matches the user's current needs and emotional state by comprehensively analyzing the user's request text, user emotion tags, and textual user profiles.

[0099] Further, step S203 includes: First, inputting the user request text and user emotion tag into the coarse-ranking unit in the radio content recommendation module to perform semantic matching on all content items in the preset radio content library, obtaining the matching degree between each content item and the user request text and user emotion tag. Specifically, the task of the coarse-ranking unit is to perform semantic matching on all content items in the preset radio content library and calculate the matching degree between each content item and the user request text and emotion tag. For example, the user request text is "play a relaxing folk song", and the emotion tag is "neutral". The coarse-ranking unit will use natural language processing technology to analyze the description text of each song or podcast in the content library, extract its semantic features, and compare them with the user request text and emotion tag. The matching degree calculation may be based on cosine similarity or other semantic similarity algorithms to obtain the matching score between each content item and the user request and emotion.

[0100] Next, content items with a matching degree higher than a preset matching threshold are selected to form a coarse-grained candidate list. Specifically, the preset matching threshold is a filtering parameter that ensures that the content items entering the candidate list are semantically highly relevant to user needs and sentiment tags. For example, if the preset matching threshold is 0.6, only content items with a matching degree higher than 0.6 will be selected into the coarse-grained candidate list. This filtering process greatly reduces the computational workload of subsequent fine-grained ranking while ensuring the quality of candidate content.

[0101] Next, the textualized user profile and the coarse-ranked candidate list are input into the fine-ranking unit of the radio content recommendation module for calculation. This yields the profile fit and intent matching score for each content item in the coarse-ranked candidate list, based on the interest and preference features extracted from the textualized user profile by the fine-ranking unit. Specifically, the task of the fine-ranking unit is to further evaluate the profile fit and intent matching score of each candidate content item with the user's long-term interests and preferences. Fit primarily measures the degree of match between the content item and the user's interests and preferences. For example, if the user profile shows a high interest in classical music, then classical music programs will score higher in fit. The intent matching score considers the user's current intent and needs. For example, a user might want to listen to relaxing music at a specific time, so music content items related to relaxation will receive a higher intent matching score. The fine-ranking unit extracts interest and preference features from the textualized user profile. For example, if the user's interests are "photography, hiking, independent films," and the preferred genres are "science fiction, folk, documentary podcasts, classical music." By calculating the similarity between each candidate content item and these interest and preference features, a fit score is obtained for each content item. This process not only considers the user's current needs and emotions but also incorporates the user's long-term interests, making recommendations more personalized.

[0102] Then, the candidate items are comprehensively sorted in descending order based on a weighted average of the profile matching score and intent matching score to obtain the refined ranking result. Specifically, a comprehensive score is obtained by weighting the profile matching score and intent matching score of each content item. The weights can be adjusted according to actual needs and user behavior data to ensure the balance and adaptability of the recommendation results. For example, if the system finds that users are more inclined to meet immediate needs at a certain time, the weight of the intent matching score can be appropriately increased. Then, the candidate items are sorted in descending order according to this comprehensive score to obtain the final refined ranking result. This sorting method ensures that the content items in the recommendation list not only highly match the user's long-term interests but also meet the user's current specific needs, thereby providing a more personalized and accurate recommendation experience.

[0103] Finally, a predetermined number of content items are selected from the refined ranking results to determine the recommended radio station content. Specifically, the predetermined number is a system parameter that determines the length of the recommendation list. For example, if the predetermined number is 5, the system will select the top 5 content items from the refined ranking results as the final recommended radio station content. These content items not only highly match the user's needs and emotional tags semantically, but also closely align with the user's interests and preferences, thus providing users with highly personalized radio station content that resonates strongly with their emotions.

[0104] For example, suppose the user is a 25-year-old male with an INFP personality type, interests in photography, hiking, and independent films, and preferred genres such as science fiction, folk, documentary podcasts, and classical music. The user's current input request text is "Play a relaxing folk song," and the mood tag is "neutral." Through the above recommendation process, the system might ultimately recommend the following:

[0105] Chen Hongyu's "Ideal Thirty Years" (folk song, lighthearted).

[0106] Pu Shu's "Those Flowers" (folk song, lighthearted).

[0107] National Geographic podcast (documentary, related to the user's interest preference "hiking").

[0108] The Lonely Planet podcast (travel, related to the user's interest preference "hiking").

[0109] The "Photography World" podcast (related to the user's interest preference "photography").

[0110] Through this series of meticulous and coherent processing steps, the radio content recommendation module can fully utilize users' interaction data, emotional state, and long-term interests to generate highly personalized radio content that matches the user's current emotional state.

[0111] Step S204: Integrate the textual user profile, user sentiment tags, current query information, and recommended radio content to obtain user instructions.

[0112] It should be noted that this process ensures that the generated content not only aligns with users' long-term preferences and current mood, but also accurately meets their immediate needs.

[0113] Further, step S204 includes: First, determining the integration order of textual user profiles, user sentiment tags, current query information, and recommended radio content. Specifically, the integration order is determined based on the importance and logical relationships of the information. Generally, current query information represents the user's most direct need and should therefore be placed first; second is the textual user profile, which provides the user's long-term preferences and is the basis for generating personalized content; user sentiment tags reflect the user's current emotional state and have a significant impact on the emotional tone of the content, therefore they are ranked third; finally, recommended radio content is the specific recommendations generated by the system based on the above information, serving as the final part of the integration.

[0114] Next, the current query information, textualized user profile, user sentiment tags, and recommended radio content are sequentially concatenated to form the initial instruction text, following the integration order. Specifically, assuming the current query information is "play a relaxing folk song," the textualized user profile is "25-year-old male, mild-mannered, loves photography and hiking," the user sentiment tag is "neutral," and the recommended radio content is "Chen Hongyu's 'Ideal Thirty Years'," the initial instruction text might be as follows: "Play a relaxing folk song; 25-year-old male, mild-mannered, loves photography and hiking; sentiment state: neutral; recommended content: Chen Hongyu's 'Ideal Thirty Years'."

[0115] Next, duplicate information in the initial instruction text is removed to obtain the first instruction text. Specifically, the purpose of deduplication is to ensure that information in the instruction text is not repeated, thereby improving the clarity and efficiency of the instructions. For example, if the initial instruction text mentions "neutral" multiple times, it will only be retained once in the first instruction text. After deduplication, the first instruction text might look like this: "Play a relaxing folk song; 25-year-old male, mild-mannered, loves photography and hiking; emotional state: neutral; recommended content: Chen Hongyu's 'Ideal Thirty Years'."

[0116] Then, based on a preset instruction format template, an instruction identifier and a generation timestamp are added to the first instruction text to obtain the second instruction text. Specifically, the instruction identifier is used to identify the type and source of the instruction, while the timestamp records the specific time the instruction was generated, ensuring the timeliness and traceability of the instruction. For example, the preset instruction format template may require adding an "Instruction ID" at the beginning of the instruction and "Generation Time" at the end. Assuming the instruction ID is "001" and the generation time is "2025-10-15 14:30:00", the second instruction text may be as follows: "Instruction ID: 001; Play a relaxing folk song; 25-year-old male, mild-mannered, loves photography and hiking; emotional state: neutral; recommended content: Chen Hongyu's 'Ideal Thirty Years'; generation time: 2025-10-15 14:30:00".

[0117] Finally, the second instruction text undergoes an integrity check. If the second instruction text meets preset requirements, it is identified as a user instruction. Specifically, the purpose of the integrity check is to ensure that the instruction text contains all necessary information and is formatted correctly. Preset requirements may include the structural integrity of the instruction text, the accuracy of the information, etc. If the second instruction text meets these requirements, the system identifies it as the final user instruction for subsequent radio content generation and recommendation.

[0118] Step S205: Input the user command into the radio content generation module to generate personalized radio content.

[0119] It's important to note that the radio content generation module receives meticulously integrated and validated user instructions. These instructions include key information such as user request text, textualized user profiles, user sentiment tags, and recommended radio content. This information provides comprehensive guidance to the generation module, enabling it to produce highly personalized radio content that aligns with the user's current emotional state.

[0120] Further, step S205 includes: First, the user instruction is input into the radio content generation module for semantic parsing to obtain core demand information. Specifically, the user instruction is a structured text that integrates user needs, emotion tags, profile information, and recommended content. The radio content generation module uses natural language processing technology to perform deep parsing of the user instruction and extract the user's core needs. For example, the user instruction might be: "Play a relaxing folk song; 25-year-old male, mild-mannered, loves photography and hiking; emotional state: neutral; recommended content: Chen Hongyu's 'Ideal Thirty Years'". Through semantic parsing, the module can extract core demand information, such as "play a relaxing folk song" and "recommend Chen Hongyu's 'Ideal Thirty Years'".

[0121] Subsequently, the core requirements information is matched against a pre-defined content generation rule base to determine the corresponding content generation strategy. This strategy includes tone and content structure. Specifically, the content generation rule base is a collection of multiple content generation strategies, each defining a specific tone and content structure. For example, for the "neutral" emotion tag, the rule base might specify a warm, healing tone and a relaxed, natural content structure. By matching the core requirements information with the rule base, the module determines the generation strategy, providing guidance for subsequent content generation.

[0122] Then, based on the content generation strategy, the user instruction is sent as an input prompt to the large language model to obtain the initial radio content text. Specifically, the large language model (such as GPT or other similar models) is a powerful natural language generation tool capable of generating coherent and natural text based on input prompts. After semantic parsing and strategy matching, the user instruction is transformed into a specific input prompt and sent to the large language model. For example, the input prompt might be: "You are a professional and empathetic radio host. Please generate a natural, fluent, and context-appropriate dialogue based on the following user profile, user sentiment, recommended content, and user request. Play a light folk song; 25-year-old male, mild-mannered, loves photography and hiking; emotional state: neutral; recommended content: Chen Hongyu's 'Ideal Thirty Years'." The large language model generates an initial radio content text based on the input prompt and the preset content generation strategy.

[0123] Finally, the initial radio content text is formatted and optimized to create personalized radio content. Specifically, the purpose of format optimization is to ensure that the generated radio content conforms to radio playback standards in terms of format, while also being more closely aligned with the user's personalized needs in terms of content. The optimization process may include adjusting the text length, improving the fluency of the language, and adding appropriate transition sentences. For example, the initial radio content text might be: "Dear listeners, today we're playing a very relaxing folk song—Chen Hongyu's 'Ideal Thirty Years.' This song is perfect for listening to on a quiet afternoon, allowing your mind to relax." After format optimization, the final personalized radio content might be: "Dear listeners, welcome to today's radio time. Today, we bring you a warm and relaxing folk song—Chen Hongyu's 'Ideal Thirty Years.' We hope this song can be like a ray of warm sunshine, illuminating your mood at this moment. On this quiet afternoon, let's immerse ourselves in the melody of this song and feel that tranquility and beauty." Through these processing steps, the information in the user's instructions can be fully utilized to generate highly personalized radio content with strong emotional resonance, significantly improving the user experience and the system's intelligence level.

[0124] This embodiment generates a textualized user profile through a user profile building module and parses the user's current query information to obtain the required text; the emotion perception and fusion module combines interaction data and real-time environmental data to generate user emotion tags; the radio content recommendation module sorts the user's needs, emotion tags, and profile to obtain recommended content; the above information is integrated to form a user command, which is input into the radio content generation module to generate personalized radio content. This can accurately capture user needs and emotions, and combine them with the real-time environment to generate highly personalized and contextualized radio content, significantly improving user experience and system intelligence.

[0125] Based on the first embodiment of this application, this application also provides a radio content generation device based on user profiles and emotion perception. Please refer to... Figure 5 The device includes:

[0126] The acquisition module 10 is used to acquire user data, which includes user profile data, interaction data, real-time environment data, and current query information.

[0127] Result module 20 is used to process user data input into a preset radio content generation model to obtain personalized radio content. The preset radio content generation model includes a user profile construction module, an emotion perception and fusion module, a radio content recommendation module, and a radio content generation module. It is also used to construct a textual user profile from the user profile construction module and parse the current query information to obtain the user's request text; to perform emotion recognition on the interaction data through the emotion perception and fusion module and fuse it with the real-time environmental data to obtain user emotion tags; to sort the user request text, user emotion tags, and textual user profile through the radio content recommendation module to obtain recommended radio content; to integrate the textual user profile, user emotion tags, current query information, and recommended radio content to obtain user instructions; and to input the user instructions into the radio content generation module to generate personalized radio content.

[0128] The radio content generation device based on user profiles and emotion perception provided in this application, employing the radio content generation method based on user profiles and emotion perception described in the above embodiments, can solve the technical problem of how to generate radio content that better meets users' personalized needs through user profiles and emotion perception. Compared with the prior art, the beneficial effects of the radio content generation device based on user profiles and emotion perception provided in this application are the same as those of the radio content generation method based on user profiles and emotion perception provided in the above embodiments, and other technical features in the radio content generation device based on user profiles and emotion perception are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0129] In one embodiment, the acquisition module 10 is further configured to: send a data retrieval instruction to a user profile database, so that the user profile database returns profile data, wherein the profile data includes user identifier, age, gender, interests and preferences, and personality type; collect interaction data generated during the user's interaction with the radio station, wherein the interaction data includes text data and voice interaction data input by the user; send an environmental data acquisition instruction to a third-party environmental service platform, so that the third-party environmental service platform returns real-time weather data and real-time time period data of the user's current location, wherein the real-time environmental data includes the real-time weather data and the real-time time period data; receive query content input by the user through the radio interaction interface, and determine the query content as the current query information; and perform format verification and integration on the profile data, the interaction data, the real-time environmental data, and the current query information to form user data.

[0130] In one embodiment, the result module 20 is further configured to input the profile data into the user profile construction module for processing to obtain key fields in the profile data, wherein the key fields include age, gender, interests and preferences, and personality type; fill the key fields into a preset profile generation template to generate initial user profile text, wherein the profile generation template is constructed by the user profile construction module; perform semantic optimization on the initial user profile text to obtain a textualized user profile; input the current query information into the user profile construction module for parsing to obtain the core intent and key entities in the current query information, wherein the core intent includes play, search, and query; and organize the core intent and the key entities according to a preset structured format to obtain user demand text.

[0131] In one embodiment, the result module 20 is further configured to: input the voice interaction data in the interaction data into the voice emotion recognition unit in the emotion perception and fusion module for feature extraction and emotion classification, and output a first emotion probability vector; input the text data in the interaction data into the text emotion classification unit in the emotion perception and fusion module for semantic analysis and emotion judgment, and output a second emotion probability vector; select a target emotion probability vector from the first emotion probability vector and the second emotion probability vector as the user emotion vector according to a preset priority rule; match the real-time weather data and real-time time period data in the real-time environment data according to a preset environmental emotion mapping rule library to obtain a weather emotion vector and a time emotion vector respectively; calculate the weather emotion vector and the time emotion vector to obtain an environmental emotion vector; perform weighted fusion of the user emotion vector and the environmental vector to obtain a fused emotion vector; sort the fused emotion vectors by probability and select the fused emotion vector with the highest probability as the user emotion label.

[0132] In one embodiment, the result module 20 is further configured to input the user demand text and the user emotion tag into the coarse ranking unit in the radio content recommendation module to perform semantic matching on all content items in the preset radio content library, and obtain the matching degree of each content item with the user demand text and the user emotion tag; and select content items with a matching degree higher than a preset matching threshold to form a coarse ranking candidate list.

[0133] The textualized user profile and the coarse-ranked candidate list are input into the fine-ranking unit of the radio content recommendation module for calculation. This yields a profile matching score and intent matching score for each content item in the coarse-ranked candidate list, based on the interest and preference features extracted from the textualized user profile by the fine-ranking unit. The candidate items are then weighted and sorted in descending order based on the profile matching score and intent matching score to obtain the fine-ranking result. A preset number of content items are selected from the fine-ranking result to determine the recommended radio content.

[0134] In one embodiment, the result module 20 is further configured to determine the integration order of the textual user profile, the user emotion tag, the current query information, and the recommended radio content; sequentially concatenate the current query information, the textual user profile, the user emotion tag, and the recommended radio content according to the integration order to form an initial instruction text; perform deduplication on the duplicate information in the initial instruction text to obtain a first instruction text; add an instruction identifier and generate a timestamp to the first instruction text based on a preset instruction format template to obtain a second instruction text; perform integrity verification on the second instruction text, and when the second instruction text meets preset requirements, determine the second instruction text as a user instruction.

[0135] In one embodiment, the result module 20 is further configured to input the user instruction into the radio content generation module for semantic parsing to obtain core requirement information; match the core requirement information with a preset content generation rule base to determine the corresponding content generation strategy, wherein the content generation strategy includes tone style and content structure; send the user instruction as an input prompt to the large language model according to the content generation strategy to obtain initial radio content text; and optimize the format of the initial radio content text to determine it as personalized radio content.

[0136] This application provides a radio content generation device based on user profiles and emotion perception. The radio content generation device based on user profiles and emotion perception includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the radio content generation method based on user profiles and emotion perception in the above embodiment 1.

[0137] The following is for reference. Figure 6 The diagram illustrates a structural schematic of a radio content generation device based on user profiles and emotion perception, suitable for implementing embodiments of this application. The radio content generation device based on user profiles and emotion perception in the embodiments of this application may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 6 The radio content generation device based on user profiles and emotion perception shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0138] like Figure 6As shown, the radio content generation device based on user profiles and emotion perception may include a processing unit 1001 (e.g., a central processing unit, a graphics processor, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1002 or a program loaded from a storage device 1003 into a random access memory (RAM) 1004. The RAM 1004 also stores various programs and data required for the operation of the radio content generation device based on user profiles and emotion perception. The processing unit 1001, ROM 1002, and RAM 1004 are interconnected via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following can be connected to I / O interface 1006: input devices 1007 including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices 1008 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 1003 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1009. Communication device 1009 allows the user-profile and emotion-aware radio content generation device to communicate wirelessly or wiredly with other devices to exchange data. Although various user-profile and emotion-aware radio content generation devices are shown in the figures, it should be understood that it is not required to implement or possess all of them. More or fewer may be implemented alternatively.

[0139] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable storage medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from ROM 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0140] The radio content generation device based on user profiles and emotion perception provided in this application, employing the radio content generation method based on user profiles and emotion perception described in the above embodiments, can solve the technical problem of how to generate radio content that better meets users' personalized needs through user profiles and emotion perception. Compared with the prior art, the beneficial effects of the radio content generation device based on user profiles and emotion perception provided in this application are the same as those of the radio content generation method based on user profiles and emotion perception provided in the above embodiments, and other technical features in this radio content generation device based on user profiles and emotion perception are the same as those disclosed in the previous embodiment method, and will not be repeated here.

[0141] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0142] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0143] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, the computer-readable program instructions being used to execute the radio content generation method based on user profiles and emotion perception in the above embodiments.

[0144] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections with one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible storage medium containing or storing a program that can be executed by instructions, used by a device, or used in conjunction with it. The program code contained on the computer-readable storage medium may be transmitted using any suitable storage medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0145] The aforementioned computer-readable storage medium may be included in a radio content generation device based on user profiling and emotion perception; or it may exist independently and not assembled into a radio content generation device based on user profiling and emotion perception.

[0146] The aforementioned computer-readable storage medium carries one or more programs that, when executed by a radio content generation device based on user profiling and emotion perception, enable the device to write computer program code for performing the operations of this application in one or more programming languages ​​or a combination thereof. These programming languages ​​include object-oriented programming languages—such as Java, Smalltalk, and C++—and conventional procedural programming languages—such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0147] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of methods and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using dedicated hardware-based implementations that perform the specified functions or operations, or can be implemented using a combination of dedicated hardware and computer instructions.

[0148] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0149] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the above-described radio content generation method based on user profiles and emotion perception. This solves the technical problem of how to generate radio content that better meets users' personalized needs through user profiles and emotion perception. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the radio content generation method based on user profiles and emotion perception provided in the above embodiments, and will not be repeated here.

[0150] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the radio content generation method based on user profiles and emotion perception as described above.

[0151] The computer program product provided in this application solves the technical problem of how to generate radio content that better meets users' personalized needs through user profiling and emotion perception. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as those of the radio content generation method based on user profiling and emotion perception provided in the above embodiments, and will not be repeated here.

[0152] The above description is only a part of the embodiments of this application and does not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A method for generating radio content based on user profiles and emotion perception, characterized in that, include: Acquire user data, wherein the user data includes profile data, interaction data, real-time environmental data, and current query information; The user data is input into a preset radio content generation model for processing to obtain personalized radio content. The preset radio content generation model includes a user profile construction module, an emotion perception and fusion module, a radio content recommendation module, and a radio content generation module. The step of inputting the user data into a preset radio content generation model for processing to obtain personalized radio content includes: The user profile data is used to construct a textual user profile through the user profile construction module, and the current query information is parsed to obtain the user's request text. The user's emotion tag is obtained by performing emotion recognition on the interaction data through the emotion perception and fusion module and then fusing and reconstructing it with the real-time environmental data. The radio content recommendation module sorts the user request text, the user emotion tags, and the textual user profile to obtain recommended radio content. The user command is obtained by integrating the textualized user profile, the user emotion tag, the current query information, and the recommended radio content. The user command is input into the radio content generation module to generate personalized radio content; The step of performing emotion recognition on the interaction data through the emotion perception and fusion module and fusing and reconstructing it with the real-time environmental data to obtain the user's emotion tag includes: The voice interaction data in the interaction data is input into the voice emotion recognition unit in the emotion perception and fusion module for feature extraction and emotion classification, and the first emotion probability vector is output. The text data in the interaction data is input into the text sentiment classification unit in the emotion perception and fusion module for semantic analysis and sentiment judgment, and a second emotion probability vector is output. According to a preset priority rule, a target emotion probability vector is selected from the first emotion probability vector and the second emotion probability vector as the user emotion vector; Based on a preset environmental emotion mapping rule base, the real-time weather data and real-time time period data in the real-time environmental data are matched to obtain weather emotion vectors and time emotion vectors respectively. The environmental emotion vector is obtained by calculating the weather emotion vector and the time emotion vector. The user emotion vector and the environmental emotion vector are weighted and fused to obtain a fused emotion vector. The fused emotion vectors are sorted by probability, and the fused emotion vector with the highest probability is selected as the user's emotion label.

2. The method as described in claim 1, characterized in that, The steps for obtaining user data include: Send a data retrieval command to the user profile database so that the user profile database returns profile data, wherein the profile data includes user identifier, age, gender, interests and preferences and personality type; Collect interaction data generated during the interaction between the user and the radio station, including text data input by the user and voice interaction data; Send an environmental data acquisition instruction to a third-party environmental service platform so that the third-party environmental service platform can return real-time weather data and real-time time period data of the user's current location, wherein the real-time environmental data includes the real-time weather data and the real-time time period data; Receive the query content input by the user through the radio interactive interface, and determine the query content as the current query information; The user profile data, the interaction data, the real-time environment data, and the current query information are format-validated and integrated to form user data.

3. The method as described in claim 1, characterized in that, The steps of constructing a textualized user profile from the profile data using the user profile construction module and parsing the current query information to obtain the user's request text include: The profile data is input into the user profile building module for processing to obtain key fields in the profile data, wherein the key fields include age, gender, interests and preferences, and personality type; The key fields are filled into the preset profile generation template to generate the initial user profile text, wherein the profile generation template is constructed by the user profile construction module. The initial user profile text is semantically optimized to obtain a textualized user profile; The current query information is input into the user profile building module for parsing to obtain the core intent and key entities in the current query information, wherein the core intent includes play, search, and query. The core intent and key entities are organized according to a preset structured format to obtain the user requirement text.

4. The method as described in claim 1, characterized in that, The step of sorting the user request text, the user emotion tag, and the textual user profile through the radio content recommendation module to obtain recommended radio content includes: The user request text and the user emotion tag are input into the coarse ranking unit in the radio content recommendation module to perform semantic matching on all content items in the preset radio content library, and obtain the matching degree between each content item and the user request text and the user emotion tag; Select content items with a matching degree higher than a preset matching threshold to form a coarse candidate list; The textual user profile and the coarse-ranked candidate list are input into the fine-ranking unit of the radio content recommendation module for calculation, to obtain the profile matching score and intent matching score of each content item in the coarse-ranked candidate list with interest and preference features, wherein the interest and preference features are extracted from the textual user profile by the fine-ranking unit; Based on the weighted average of the profile matching score and intent matching score, the candidates are comprehensively sorted in descending order to obtain the refined ranking result; Select a preset number of content items from the ranking results to determine the recommended radio station content.

5. The method as described in claim 1, characterized in that, The step of integrating the textualized user profile, the user emotion tag, the current query information, and the recommended radio content to obtain the user instruction includes: Determine the integration order of the textualized user profile, the user sentiment tags, the current query information, and the recommended radio content; The current query information, the textualized user profile, the user emotion tags, and the recommended radio content are sequentially concatenated according to the integration order to form the initial instruction text; The duplicate information in the initial instruction text is deduplicated to obtain the first instruction text; Based on a preset instruction format template, an instruction identifier and a timestamp are added to the first instruction text to obtain the second instruction text; The second instruction text is subjected to integrity verification. When the second instruction text meets the preset requirements, the second instruction text is determined to be a user instruction.

6. The method as described in claim 1, characterized in that, The step of inputting the user instruction into the radio content generation module to generate personalized radio content includes: The user command is input into the radio content generation module for semantic parsing to obtain core requirement information; The core requirement information is matched with a preset content generation rule base to determine the corresponding content generation strategy, wherein the content generation strategy includes tone and content structure. According to the content generation strategy, the user instruction is sent as an input prompt to the large language model to obtain the initial radio content text; The initial radio content text is formatted and optimized to determine personalized radio content.

7. A radio content generation device based on user profiles and emotion perception, characterized in that, The device includes: The acquisition module is used to acquire user data, which includes user profile data, interaction data, real-time environment data, and current query information; The result module is used to process the user data into a preset radio content generation model to obtain personalized radio content. The preset radio content generation model includes a user profile construction module, an emotion perception and fusion module, a radio content recommendation module, and a radio content generation module. It is also used to construct a textual user profile from the profile data through the user profile construction module and parse the current query information to obtain user request text; to perform emotion recognition on the interaction data through the emotion perception and fusion module and fuse it with the real-time environmental data to obtain user emotion tags; to sort the user request text, user emotion tags, and textual user profile through the radio content recommendation module to obtain recommended radio content; to integrate the textual user profile, user emotion tags, current query information, and recommended radio content to obtain user instructions; and to input the user instructions into the radio content generation module to generate personalized radio content. It is also used to... The voice interaction data in the interaction data is input into the voice emotion recognition unit of the emotion perception and fusion module for feature extraction and emotion classification, outputting a first emotion probability vector; the text data in the interaction data is input into the text emotion classification unit of the emotion perception and fusion module for semantic analysis and emotion judgment, outputting a second emotion probability vector; a target emotion probability vector is selected from the first and second emotion probability vectors as the user emotion vector according to a preset priority rule; based on a preset environmental emotion mapping rule library, real-time weather data and real-time time period data in the real-time environmental data are matched to obtain a weather emotion vector and a time emotion vector respectively; the weather emotion vector and the time emotion vector are calculated to obtain an environmental emotion vector; the user emotion vector and the environmental emotion vector are weighted and fused to obtain a fused emotion vector; the fused emotion vectors are sorted by probability, and the fused emotion vector with the highest probability is selected as the user emotion label.

8. A radio content generation device based on user profiles and emotion perception, characterized in that, The device includes: a memory, a processor, and a radio content generation program based on user profiles and emotion perception stored in the memory and running on the processor, the radio content generation program based on user profiles and emotion perception being configured to implement the steps of the radio content generation method based on user profiles and emotion perception as described in any one of claims 1-6.

9. A storage medium, characterized in that, The storage medium stores a radio content generation program based on user profiles and emotion perception. When the radio content generation program based on user profiles and emotion perception is executed by the processor, it implements the steps of the radio content generation method based on user profiles and emotion perception as described in any one of claims 1-6.

Images