Resource recommendation method and system based on large model, terminal and storage medium

Abstract

The invention discloses a resource recommendation method and system based on a large model, a terminal and a storage medium. The method comprises: acquiring voice data of a user, and determining a user intention according to the voice data; judging whether the user intention is a large model question and answer intention; if the user intention is a large model question and answer intention, obtaining historical interaction information of the user, and generating large model question and answer data according to the historical interaction information and the voice data; and inputting the question and answer data into the large model to obtain resource recommendation information output by the large model. According to the method, the voice data of the user and the historical interaction information of the user are combined to generate the question and answer data of the large model. Since the historical interaction information can cover past search records and interaction behaviors of the user, the real demand of the user can be deeply understood through the question and answer data jointly generated by the historical interaction information and the sound information, so that resources, such as music resources or movie and television resources, which accord with interests and preferences of the user can be accurately recommended to the user in a personalized manner.

Description

Technical Field

[0001] The present invention relates to the field of intelligent system technology, and in particular to a resource recommendation method, system, terminal and storage medium based on a large model. Background Art

[0002] With the popularization of smart home appliances, home appliances are interconnected, and users interact with home appliances more frequently. The existing home appliance recommendation system has a relatively simple operation mode, which only provides pure text answers based on a single question input by the user. This interactive method lacks in-depth analysis of the user's input data, making it difficult to accurately push content recommendation information that matches the user's interests and preferences, and unable to meet the diverse needs of users, resulting in limited depth of interaction between users and home appliances.

[0003] Therefore, the existing technology still needs to be improved and developed. Summary of the invention

[0004] The technical problem to be solved by the present invention is that, in response to the above-mentioned defects of the prior art, a resource recommendation method, system, terminal and storage medium based on a large model are provided, aiming to solve the problem that the recommendation system of home appliance products in the prior art lacks in-depth analysis of user input information, making it difficult to accurately push content recommendation information that matches the user's interests and preferences.

[0005] The technical solution adopted by the present invention to solve the problem is as follows:

[0006] In a first aspect, an embodiment of the present invention provides a resource recommendation method based on a large model, the method comprising:

[0007] Acquire voice data of the user, and determine the user's intention based on the voice data;

[0008] Determine whether the user intention is a large model question-answering intention;

[0009] If the user intention is a large-model question-and-answer intention, obtaining the user's historical interaction information, and generating large-model question-and-answer data according to the historical interaction information and the voice data;

[0010] The question and answer data is input into the big model to obtain resource recommendation information output by the big model.

[0011] In one embodiment, the method further comprises:

[0012] If the user intention is not a large model question and answer intention, the corresponding application control instruction is executed according to the voice data.

[0013] In one embodiment, generating the question-answer data of the large model according to the historical interaction information and the voice data includes:

[0014] Generate a prompt word according to the historical interaction information; or generate a prompt word according to the historical interaction information and the user's voice information;

[0015] A large model of question and answer data is generated based on the voice data and the prompt words.

[0016] In one implementation, the historical interaction information includes context information and historical behavior information of the user; and the sound information includes voiceprint information of the user.

[0017] In one implementation, the question-answer data is input into the big model to obtain resource recommendation information output by the big model, and then the following steps are further included:

[0018] Determining whether the resource recommendation information includes a target noun related to the target media posture;

[0019] If the target noun is included, the target media asset is called to obtain the resource corresponding to the target noun to obtain the target resource.

[0020] In one implementation, after obtaining the target resource, the method further includes:

[0021] The target resource is pushed to the terminal device for display and / or playback.

[0022] In one implementation, the target media assets include film and television media and / or music media.

[0023] In a second aspect, an embodiment of the present invention further provides a resource recommendation system based on a large model, the system comprising:

[0024] A speech analysis module, used to obtain the user's speech data and determine the user's intention based on the speech data;

[0025] An intention judgment module is used to judge whether the user's intention is a large model question-answering intention;

[0026] A question-and-answer generation module, configured to obtain historical interaction information of the user if the user intention is a large-model question-and-answer intention, and generate large-model question-and-answer data according to the historical interaction information and the voice data;

[0027] The big model recommendation module is used to input the question and answer data into the big model to obtain resource recommendation information output by the big model.

[0028] In a third aspect, an embodiment of the present invention further provides a terminal, comprising a memory and one or more processors; the memory stores one or more programs; the program comprises instructions for executing any of the large model-based resource recommendation methods described above; and the processor is used to execute the program.

[0029] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium on which a plurality of instructions are stored, wherein the instructions are suitable for being loaded and executed by a processor to implement the steps of any of the large model-based resource recommendation methods described above.

[0030] Beneficial effects of the present invention: The embodiments of the present invention obtain the user's voice data, determine the user's intention based on the voice data; determine whether the user's intention is a large model question and answer intention; if the user's intention is a large model question and answer intention, obtain the user's historical interaction information, generate large model question and answer data based on the historical interaction information and the voice data; input the question and answer data into the large model to obtain resource recommendation information output by the large model. The present invention combines the user's own voice data with the user's historical interaction information to generate large model question and answer data. Since historical interaction information can cover the user's past search records and interaction behaviors, the question and answer data generated by historical interaction information and sound information can deeply understand the user's real needs, thereby accurately recommending resources that match their interests and preferences to the user in a personalized manner, such as music resources or film and television resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0032] Figure 1 It is a flowchart of a resource recommendation method based on a large model provided in an embodiment of the present invention.

[0033] Figure 2 It is a complete flow chart of the resource recommendation method based on a large model provided by an embodiment of the present invention.

[0034] Figure 3 It is a module diagram of a resource recommendation system based on a large model provided in an embodiment of the present invention.

[0035] Figure 4 It is a principle block diagram of a terminal provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0036] The present invention discloses a resource recommendation method, system, terminal and storage medium based on a large model. In order to make the purpose, technical solution and effect of the present invention clearer and more specific, the present invention is further described in detail with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0037] It will be understood by those skilled in the art that, unless expressly stated, the singular forms "one", "said", and "the" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present invention refers to the presence of the features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof. It should be understood that when we refer to an element as being "connected" or "coupled" to another element, it may be directly connected or coupled to the other element, or there may be intermediate elements. In addition, the "connection" or "coupling" used herein may include wireless connection or wireless coupling. The term "and / or" used herein includes all or any unit and all combinations of one or more associated listed items.

[0038] It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as generally understood by those skilled in the art in the art to which the present invention belongs. It should also be understood that terms such as those defined in general dictionaries should be understood to have meanings consistent with the meanings in the context of the prior art, and will not be interpreted with idealized or overly formal meanings unless specifically defined as herein.

[0039] In view of the above-mentioned defects of the prior art, the present invention provides a resource recommendation method based on a big model, the method obtains the user's voice data, determines the user's intention according to the voice data; determines whether the user's intention is a big model question and answer intention; if the user's intention is a big model question and answer intention, obtains the user's historical interaction information, generates the big model's question and answer data according to the historical interaction information and the voice data; inputs the question and answer data into the big model, and obtains the resource recommendation information output by the big model. The present invention combines the user's own voice data with the user's historical interaction information to generate the big model's question and answer data. Since the historical interaction information can cover the user's past search records and interaction behaviors, the question and answer data generated by the historical interaction information and the sound information can deeply understand the user's real needs, thereby personalized and accurately recommending resources that fit their interests and preferences to the user, such as music resources or film and television resources.

[0040] like Figure 1As shown, the method specifically comprises the following steps:

[0041] Step S100: Acquire the user's voice data, and determine the user's intention based on the voice data.

[0042] Specifically, the user refers to a user of a home appliance who needs content recommendation, such as a user of a smart TV. Acquiring the user's voice data is a basic step of this embodiment. By analyzing the user's voice data, the user's intention can be preliminarily determined, and the topic direction that the user wants to ask or understand at the moment can be understood.

[0043] For example, the automatic speech recognition (ASR) technology is used to accurately capture the user's voice content, and the recognized user query is used as a request and sent to the intent recognition module for intent recognition. The present invention uses the powerful text understanding and generation capabilities of the GPT large model, combined with ASR to accurately recognize voice words, to deeply analyze user intentions. Whether it is vague daily expressions or inquiries in professional fields, it can accurately locate the needs, and then associate film, television, and music resources that are highly in line with the user's interests, significantly improving the matching degree between recommended content and user expectations.

[0044] Step S200: Determine whether the user intention is a large model question and answer intention.

[0045] Specifically, this embodiment can pre-build an intent recognition module to improve the efficiency of analyzing user intent. The intent recognition module will perform in-depth analysis of the user's voice data, determine the intent type, and then point to the corresponding processing module according to different intents, and finally feedback the corresponding instructions to the client, where the client can be an Android client. If the user's intent is determined to be a large model question and answer intent, the process will enter the large model question and answer processing module.

[0046] Step S300: If the user intention is a large-model question-and-answer intention, the user's historical interaction information is obtained, and large-model question-and-answer data is generated based on the historical interaction information and the voice data.

[0047] Specifically, when the user's intention is a big model question and answer intention, the big model question and answer processing flow is entered. The big model's question and answer data is generated through the user's voice data and historical interaction information, and a request is initiated to the big model interface. In actual application, the big model can use models such as Alitong Yi Qianwen and ChatGPT.

[0048] In one implementation, generating the question-answer data of the large model according to the historical interaction information and the voice data includes:

[0049] Generate a prompt word according to the historical interaction information; or generate a prompt word according to the historical interaction information and the user's voice information;

[0050] A large model of question and answer data is generated based on the voice data and the prompt words.

[0051] Specifically, historical interaction information may include historical conversation records of users, such as previous questions, queries, and interactive content of users. By analyzing these historical conversation records, we can understand the interests and preferences of users, and thus better analyze their needs. Historical interaction information may also include the search history of users, which can help identify the long-term interests and needs of users. Voice information may include the voice features of users, such as intonation, speaking speed, and timbre, etc. These voice features can reflect the emotions and attitudes of users, and help better understand the intentions and needs of users. Figure 2 As shown, this embodiment carefully designs prompt words through historical interaction information, or combines historical interaction information with sound information, and cleverly integrates the user's voice data into the prompt words, thereby initiating a request to the large model interface to obtain the question and answer data of the large model.

[0052] In one implementation, the historical interaction information includes context information and historical behavior information of the user; and the sound information includes voiceprint information of the user.

[0053] Specifically, this embodiment injects one or more prompt words such as the user's context information, historical behavior information, and voiceprint information into the big model. These prompt words are like precise keys. They will not only embed the user's context information, past behavior information, and voiceprint information, but also explicitly require the big model to use specific symbols to frame the names of movies and music that may be involved when returning the results. Through these prompt words, the big model can coordinate multi-dimensional data, not only can it keenly perceive the user's boredom and give intimate comfort; it will also deeply explore the user's potential preferences based on the user's past historical conversation information, unique voiceprint characteristics, and long-term accumulated behavior viewing data, so that the recommended movies, TV shows, and music works not only fit the current inquiry, but also echo the user's long-term interests and preferences, realizing intelligent and deep integration of information.

[0054] For example, the prompt words are as follows:

[0055] #Role: You are a structured information question-answering assistant. You are the big model of the voice question-answering assistant developed by XX company. Your name is Xiaowei.

[0056] #Requirements: You need to answer the user's questions, and your answers must be as brief as possible;

[0057] #Set knowledge information (important knowledge, please remember): User location information: Beijing; User location weather: Sunny; Current time information: January 9, 2025, 14:53; Thursday;

[0058] #Notes: If the user's question can be associated with the recommendation of movies, TV shows or music, please give the user a reasonable recommendation of movies, TV shows or music and answer the user's question in the shortest possible way. Do not make up non-existent movies, TV shows or music. If the question is not related to movies, TV shows or music, such as asking about date, time, week, mathematical operations, knowledge query, etc., please do not recommend questions that have nothing to do with movies, TV shows or music. If there is a recommended movie, TV show or music, only recommend one type of movie, TV show or music. At the end of the answer result, return the movie or music name in the format of the example, and return a maximum of 5. Please remember to add the "$segment" separator. For movie names, please add "《》" above the name, for example: $segment recommends these good-looking movies "xxxx" and "xxxx" for you. For music names, please add "【】" above the name, for example: $segment recommends these good-sounding music [xxxx] and [xxxx] for you. If you ask questions about actors, singers, movies, TV shows, and music, you must recommend movies, TV shows or music in the format required above. If the user's question requires the use of set knowledge information, please refer to the set knowledge information to answer. The answer must be true and accurate, and do not add fabricated elements to the answer. Please note that when answering questions about a person's information, please polish the positive aspects of the person and do not return negative or bad information. For information about the person and encyclopedia, refer to the sample template and return reasonable information based on the question. For example: name**\date of birth**\place of origin**. Answer in Chinese.

[0059] Step S400: input the question and answer data into the big model to obtain resource recommendation information output by the big model.

[0060] Specifically, the big model has been trained with a large amount of data and has strong semantic understanding and knowledge association capabilities. The previously generated question and answer data is input into the big model, and the big model can deeply process the question and answer data based on its own algorithms and knowledge reserves. For example, the question and answer data may contain a user's vague description of a certain type of film, television or music. The big model can accurately interpret it and associate it with its own knowledge of film, television and music. The resource recommendation information output by the big model (which can be called assistant content) is based on question and answer data, its pre-trained knowledge and a comprehensive judgment of user intentions. The resource recommendation information may cover various types of film, television and music resources, including popular, niche, and specific styles, etc., and can be sorted according to certain priorities.

[0061] In one implementation, the question-answer data is input into the big model to obtain resource recommendation information output by the big model, and then the method further includes:

[0062] Determining whether the resource recommendation information includes a target noun related to the target media posture;

[0063] If the target noun is included, the target media asset is called to obtain the resource corresponding to the target noun to obtain the target resource.

[0064] Specifically, after the large model returns the resource recommendation information, the system will quickly and accurately extract the target nouns related to the target media. For example, if the target media is film and television, the target noun can be the name of the film and television, and if the target media is music, the target noun can be the name of the music. Then, combined with the user's unique voiceprint information and / or user ID information, a request is sent to the interface of the target media. After receiving the request, the interface of the target media will quickly retrieve and match the corresponding resources, that is, obtain the target resources. This embodiment can deeply explore the user's potential preferences, and then recommend a series of pleasant music that fits the current mood, dispel boredom for the user, and bring pleasant auditory enjoyment, making the user feel as if he has an intimate old friend who is always by his side and understands his needs.

[0065] For example, assuming that the solution of this embodiment is applied to a voice assistant, when the user says "introduce a certain singer", the voice assistant will introduce the singer and recommend the singer's related music at the end. Or, if the user says "help me plan a three-day tour of Nanjing", the voice assistant will make a Nanjing itinerary and recommend Nanjing-related movies and TV shows. While maximizing the satisfaction of user questions, this embodiment combines the user's relevant historical conversation information, voiceprint information, and behavior viewing data to intelligently recommend film, television and music media resources.

[0066] In one implementation, after obtaining the target resource, the method further includes:

[0067] The target resource is pushed to the terminal device for display and / or playback.

[0068] Specifically, this embodiment will eventually push the target resource to the terminal device for display and playback. In this way, it is successfully achieved that while answering the user's words, relevant film, television, and music resources are intimately recommended to the user, fully satisfying the user's audio-visual needs. The entire service process of this embodiment is streamed, from voice input, intent recognition, large model question and answer to resource recommendation and playback. Users do not need to switch back and forth between multiple applications or functional modules. They can get answers to questions and favorite audio-visual resources in the same interactive session and enjoy a coherent and smooth user experience.

[0069] In one implementation, the target media assets include film and television media and / or music media.

[0070] Specifically, the target media mainly refers to audio-visual media, such as film and television media and music media. This embodiment combines the user's historical interaction information, voice data, voiceprint information, user ID and other exclusive data to create a unique film and television music recommendation list for each user. Even if different users input the same words, the recommended content they get will be different due to differences in personal historical data, truly achieving personalized service.

[0071] In one implementation, the method further includes:

[0072] If the user intention is not a large model question and answer intention, the corresponding application control instruction is executed according to the voice data.

[0073] Specifically, if the user's intention is not a large model question-and-answer intention, it means that the user may not need resource recommendations at the moment, but instead wants to control other functional modules through voice, such as turning on and off the computer, or starting other applications. Therefore, when it is determined that the user's intention is not a large model question-and-answer intention, the corresponding application control instruction is executed according to the user's voice data.

[0074] The advantages of the present invention are:

[0075] 1. The present invention comprehensively utilizes multi-dimensional data sources such as users' rich and diverse historical context records, detailed behavioral viewing data, and voiceprint information with unique recognition, to deeply explore users' potential preferences, gain insights into users' personalized needs and preferences, and change the stereotype of simple structured layout recommendations in the past. It accurately matches and recommends film, television or music media data that suits users' tastes, fully meets users' personalized audio-visual needs, allows users to easily encounter their favorite audio-visual content, and greatly improves users' efficiency and experience in discovering high-quality content.

[0076] 2. The present invention makes full use of the extraordinary personalized answering capabilities of the big model to closely link the two key links of answering user questions and intelligently recommending resources. The big model can dig deep into the deep demands behind the user's words, and make fewer mistakes and higher fit when recommending movies, TV shows, and music. For example, when users express their needs with metaphors and witty words, they can also respond accurately. When a user initiates an inquiry, not only can they get an accurate answer quickly, but also in the same interactive process, based on the big model's deep understanding of the problem and user background, they can instantly receive recommendations for high-quality audio-visual resources related to it, realize the integration of information interaction and content acquisition, and the seamless connection between questions and answers and recommendations, solve users' information exploration and content acquisition demands in one stop, enhance the interactive stickiness between users and devices, and reshape the new paradigm of intelligent interaction for home audio-visual entertainment.

[0077] 3. The present invention fully taps the potential of audio-visual equipment. In view of the drawbacks of the large model assistant and recommendation system on TV in the past, which were independent of each other and could not coordinate their advantages, the present invention is committed to integrating the advantages of all parties, so that audio-visual equipment such as TV can not only give full play to its natural resource advantages as a platform for film and television and music playback, but also use cutting-edge technology to achieve intelligent and humanized interactive recommendations, comprehensively improve the intelligent service level of the equipment, and redefine the user experience standard of the home audio-visual entertainment center.

[0078] 4. The present invention integrates multiple data such as voice, context, behavior, voiceprint, long-term behavior trajectory, etc., to comprehensively outline the user's interest portrait, and explore potential points of interest that are ignored by ordinary systems, making the recommended content richer and more diverse, and making the recommendation results vary from content to presentation order according to the individual, creating an extremely personalized experience, far exceeding the universal recommendation effect of ordinary systems.

[0079] 5. Greater interaction flexibility: Ordinary dialogue systems have stiff and stereotyped responses and lack flexibility. The streaming service process of the present invention supports dynamic interaction and can adjust the recommendation strategy in real time based on the large model response. When users ask questions or provide additional information, the system can also quickly give an adaptive response, and the interactivity far exceeds that of traditional systems.

[0080] Based on the above embodiments, the present invention also provides a resource recommendation system based on a large model, such as Figure 3 As shown, the system comprises:

[0081] The speech analysis module 01 is used to obtain the user's speech data and determine the user's intention according to the speech data;

[0082] Intention judgment module 02, used to judge whether the user intention is a large model question-answering intention;

[0083] A question-answering generation module 03 is used to obtain the user's historical interaction information if the user's intention is a large-model question-answering intention, and generate large-model question-answering data according to the historical interaction information and the voice data;

[0084] The big model recommendation module 04 is used to input the question and answer data into the big model to obtain the resource recommendation information output by the big model.

[0085] Based on the above embodiment, the present invention further provides a terminal, whose principle block diagram can be shown as follows: Figure 4As shown. The terminal includes a processor, a memory, a network interface, and a display screen connected through a system bus. Among them, the processor of the terminal is used to provide computing and control capabilities. The memory of the terminal includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the terminal is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a resource recommendation method based on a large model is implemented, and the method includes:

[0086] Acquire voice data of the user, and determine the user's intention based on the voice data;

[0087] Determine whether the user intention is a large model question-answering intention;

[0088] If the user intention is a large-model question-and-answer intention, obtaining the user's historical interaction information, and generating large-model question-and-answer data according to the historical interaction information and the voice data;

[0089] The question and answer data is input into the big model to obtain resource recommendation information output by the big model.

[0090] The display screen of the terminal may be a liquid crystal display screen or an electronic ink display screen.

[0091] Those skilled in the art will understand that Figure 4 The principle block diagram shown in the figure is only a block diagram of a partial structure related to the scheme of the present invention, and does not constitute a limitation on the terminal to which the scheme of the present invention is applied. The specific terminal may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0092] In one implementation, the memory of the terminal stores one or more programs, and is configured to be executed by one or more processors. The one or more programs include instructions for performing a resource recommendation method based on a large model, and the method includes:

[0093] Acquire voice data of the user, and determine the user's intention based on the voice data;

[0094] Determine whether the user intention is a large model question-answering intention;

[0095] If the user intention is a large-model question-and-answer intention, obtaining the user's historical interaction information, and generating large-model question-and-answer data according to the historical interaction information and the voice data;

[0096] The question and answer data is input into the big model to obtain resource recommendation information output by the big model.

[0097] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided by the present invention can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0098] In summary, the present invention discloses a resource recommendation method, system, terminal and storage medium based on a big model. The method obtains the user's voice data, determines the user's intention according to the voice data; determines whether the user's intention is a big model question and answer intention; if the user's intention is a big model question and answer intention, obtains the user's historical interaction information, generates a big model's question and answer data according to the historical interaction information and the voice data; inputs the question and answer data into the big model to obtain the resource recommendation information output by the big model. The present invention combines the user's own voice data with the user's historical interaction information to generate a big model's question and answer data. Since the historical interaction information can cover the user's past search records and interaction behaviors, the question and answer data generated by the historical interaction information and the sound information can deeply understand the user's real needs, thereby personalized and accurately recommending resources that fit their interests and preferences to the user, such as music resources or film and television resources.

[0099] It should be understood that the application of the present invention is not limited to the above examples. For ordinary technicians in this field, improvements or changes can be made based on the above description. All these improvements and changes should fall within the scope of protection of the claims attached to the present invention.

Claims

1. A resource recommendation method based on a large model, characterized in that: The method comprises: Acquire voice data of the user, and determine the user's intention based on the voice data; Determine whether the user intention is a large model question-answering intention; If the user intention is a large-model question-and-answer intention, obtaining the user's historical interaction information, and generating large-model question-and-answer data according to the historical interaction information and the voice data; The question and answer data is input into the big model to obtain resource recommendation information output by the big model.

2. The resource recommendation method based on a large model according to claim 1, characterized in that: The method further comprises: If the user intention is not a large model question and answer intention, the corresponding application control instruction is executed according to the voice data.

3. The resource recommendation method based on a large model according to claim 1, characterized in that: Generating question-answer data of a large model according to the historical interaction information and the voice data includes: Generate a prompt word according to the historical interaction information; or generate a prompt word according to the historical interaction information and the user's voice information; A large model of question and answer data is generated based on the voice data and the prompt words.

4. The resource recommendation method based on a large model according to claim 3, characterized in that: The historical interaction information includes the context information and historical behavior information of the user; the sound information includes the voiceprint information of the user.

5. The resource recommendation method based on a large model according to claim 1, characterized in that: The question-answer data is input into the big model to obtain resource recommendation information output by the big model, and then the following steps are further included: Determining whether the resource recommendation information includes a target noun related to the target media posture; If the target noun is included, the target media asset is called to obtain the resource corresponding to the target noun to obtain the target resource.

6. The resource recommendation method based on a large model according to claim 5, characterized in that: After obtaining the target resource, it also includes: The target resource is pushed to the terminal device for display and / or playback.

7. The resource recommendation method based on a large model according to claim 5, characterized in that: The target media assets include film and television media assets and / or music media assets.

8. A resource recommendation system based on a large model, characterized in that: The system comprises: A speech analysis module, used to obtain the user's speech data and determine the user's intention based on the speech data; An intention judgment module is used to judge whether the user's intention is a large model question-answering intention; A question-and-answer generation module, configured to obtain historical interaction information of the user if the user intention is a large-model question-and-answer intention, and generate large-model question-and-answer data according to the historical interaction information and the voice data; The big model recommendation module is used to input the question and answer data into the big model to obtain resource recommendation information output by the big model.

9. A terminal, characterized in that: The terminal includes a memory and one or more processors; the memory stores one or more programs; the program contains instructions for executing the large model-based resource recommendation method as described in any one of claims 1-7; and the processor is used to execute the program.

10. A computer-readable storage medium having a plurality of instructions stored thereon, characterized in that: The instructions are suitable for being loaded and executed by a processor to implement the steps of the large model-based resource recommendation method as described in any one of claims 1-7.

Images