Music recommendation method and device, computer storage medium and program product

By converting user-perceived information and user portrait data into descriptive text and embedding them into large language models, the problems of large data demand and difficult feature extraction in the existing music recommendation system are solved, personalized recommendation and user experience improvement are achieved, and the Model training process.

CN118964733APending Publication Date: 2024-11-15NIO TECH ANHUI CO LTD

Patent Information

Application Number
CN202411024697.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-26
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

The existing music recommendation system faces the problems of large data demand, difficulty in feature extraction and integration, and insufficient user experience. Especially when user characteristics are diversified and new features are introduced, it leads to high time-consuming and limited user satisfaction.

Method used

By obtaining user perception information and user portrait data, it is converted into descriptive text and embedded in a large language model, personalized music recommendations are generated, and user requests are analyzed using natural language processing algorithms, and recommendation sorting is combined with the user's current status to reduce the complexity of model training And improve the recommended fit.

Benefits of technology

It realizes personalized recommendations with higher correlation with the user's current status, simplifies the unification of feature dimensions, improves user satisfaction and experience, avoids recommendation problems during cold start of new users, and reduces the complexity of model training.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118964733A_ABST
    Figure CN118964733A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of artificial intelligence, in particular to a music recommendation method, a music recommendation device for implementing the method, a computer storage medium for implementing the method and a computer program product. The method comprises the following steps: obtaining a candidate music list and user perception information authorized to access, wherein the candidate music list comprises candidate music determined based on a user request and user portrait data; converting the candidate music list, the user perception information and the user portrait data into descriptive texts meeting natural language processing requirements; embedding the descriptive text into first prompt information, and inputting the first prompt information into a first large language model to obtain a recommended music sequence, the recommended music being a subset of the candidate music; and returning the recommended music sequence to the user.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and more specifically to a music recommendation method, a music recommendation device implementing the method, a computer storage medium implementing the method, and a computer program product. Background Art

[0002] The current music recommendation systems rely on two main technical pillars: traditional feature engineering and deep learning methods. These systems train models that can predict users' music preferences through massive amounts of detailed annotated data.

[0003] However, such systems still face several limitations that cannot be ignored, including:

[0004] 1. Huge data demand: Faced with a huge music database and diverse user characteristics, the amount of data required to achieve accurate recommendations is enormous.

[0005] 2. Challenges in feature extraction and integration: As user portraits become increasingly rich, including but not limited to age, interest preferences and other multi-dimensional features, these features often have dimensional differences, making feature extraction and integration more difficult. For example, every time a new feature is introduced, the data needs to be rebuilt to train the model, which is both time-consuming and costly.

[0006] 3. User experience needs to be improved: Traditional music recommendation models mostly focus on song recall. This "black box" recommendation method often leads to limited user acceptance and satisfaction with the recommendation results, making it difficult to build a strong emotional connection and trust.

[0007] It should be noted that the information disclosed in the above background technology section is only used to enhance the understanding of the background of the present application, and therefore may include information that does not constitute the prior art known to ordinary technicians in the field. Summary of the invention

[0008] In order to solve or at least alleviate one or more of the above problems, the following technical solutions are provided. The embodiments of the present application provide a music recommendation method, a music recommendation device implementing the method, a computer storage medium implementing the method, and a computer program product, which can provide users with personalized recommendations that are more relevant to the user's current state, and at the same time solve the pain point of difficult feature extraction in the prior art by unifying feature dimensions.

[0009] According to a first aspect of the present application, a music recommendation method is provided, the method comprising the following steps: obtaining a candidate music list and user perception information authorized for access, wherein the candidate music list includes candidate music determined based on user requests and user portrait data; converting the candidate music list, the user perception information and the user portrait data into descriptive text that meets the requirements of natural language processing; embedding the descriptive text into a first prompt information and inputting it into a first large language model to obtain a recommended music ranking, wherein the recommended music is a subset of the candidate music; and returning the recommended music ranking to the user.

[0010] As an alternative or supplement to the above scheme, in a music recommendation method according to an embodiment of the present application, the user perception information includes one or more of the following items: user's current location, time, weather, activity scene, social status, emotional state, function usage scenario, and calendar information.

[0011] As an alternative or supplement to the above scheme, in a music recommendation method according to an embodiment of the present application, the descriptive text is embedded in the first prompt information and input into the first large language model to obtain the recommended music ranking, including: selecting recommended music from the candidate music based on the degree of association between the candidate music and the user perception information and the user portrait data; and sorting the recommended music based on the degree of association to determine the recommended music ranking.

[0012] As an alternative or supplement to the above scheme, in a music recommendation method according to an embodiment of the present application, the method also includes: parsing user requests using a natural language processing algorithm to determine the user request semantics; determining a recall algorithm that matches the user request semantics; and selecting candidate music that matches the user request semantics and the user portrait data from a music database using the recall algorithm to generate the candidate music list.

[0013] As an alternative or supplement to the above scheme, in a music recommendation method according to an embodiment of the present application, the method also includes: converting the recommended music ranking into a descriptive text that meets the requirements of natural language processing; embedding the recommended music ranking and the descriptive text of the user perception information into a second prompt information, and inputting it into a second largest language model to select the music with the highest degree of correlation with the user perception information from the recommended music; and placing the music with the highest degree of correlation as the first in the recommended music ranking.

[0014] As an alternative or supplement to the above scheme, in a music recommendation method according to an embodiment of the present application, the following operations are performed in the process of embedding the descriptive text of the user perception information into the first prompt information and / or the second prompt information: determining the weight of each type of user perception information; based on the determined weight, using a weighted sampling algorithm to filter the descriptive text of each type of user perception information; embedding the filtered descriptive text of the user perception information into the first prompt information and / or the second prompt information.

[0015] As an alternative or supplement to the above scheme, in a music recommendation method according to an embodiment of the present application, the method also includes: determining music to be played, wherein the music to be played is the music ranked first in the recommended music ranking, or the music requested by the user; embedding the introduction information of the music to be played and the descriptive text of the user perception information into a third prompt information, and inputting it into a third language model to obtain a recommendation for the music to be played; and presenting the recommendation in an interactive manner.

[0016] As an alternative or supplement to the above solution, in a music recommendation method according to an embodiment of the present application, the method further includes: randomly selecting a prompt information template from a set of prompt information templates with different sentence patterns as the third prompt information.

[0017] As an alternative or supplement to the above scheme, in a music recommendation method according to an embodiment of the present application, the method also includes: determining one or more of the image information, voice information, expression information and action information of the client's virtual image based on the recommendation language of the music to be played; and controlling the client to interact based on one or more of the image information, the voice information, the expression information and the action information.

[0018] According to a second aspect of the present application, a music recommendation device is provided, comprising: a memory; a processor; and a computer program stored on the memory and executable on the processor, wherein the execution of the computer program enables any one of the music recommendation methods described in the first aspect of the present application to be executed.

[0019] According to a third aspect of the present application, a computer storage medium is provided, wherein the computer storage medium includes instructions, and the instructions, when run, execute any one of the music recommendation methods described in the first aspect of the present application.

[0020] According to a fourth aspect of the present application, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements any one of the music recommendation methods described in the first aspect of the present application.

[0021] The music recommendation solution according to one or more embodiments of the present application actively obtains user perception information that is authorized to access, and incorporates the user perception information into the consideration of screening recommended music, so that the recommendation results are more in line with the user's current actual needs and personalized preferences, thereby improving user satisfaction and overall experience, and avoiding the disadvantage that it is difficult to accurately recommend music when new users are cold-started. In addition, in the process of screening and sorting recommended music, the solution converts all input information (that is, music lists, user perception information, and user portrait data) into descriptive text that meets the requirements of natural language processing. This conversion makes it simple to add new user features, because any new features can be converted into descriptive text and input into the model without changing the input layer structure of the model, and reduces the complexity of model training. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The above and / or other aspects and advantages of the present application will become clearer and easier to understand through the following description of various aspects in conjunction with the accompanying drawings, in which the same or similar units are represented by the same reference numerals. In the accompanying drawings:

[0023] Figure 1 is a schematic flow chart of a music recommendation method 10 according to one or more embodiments of the present application;

[0024] Figure 2 is a schematic flow chart of a music recommendation method 20 according to one or more embodiments of the present application;

[0025] Figure 3 is a schematic block diagram of a music recommendation device 30 according to one or more embodiments of the present application. DETAILED DESCRIPTION

[0026] The description of the following specific embodiments is merely exemplary in nature and is not intended to limit the disclosed technology or the application and use of the disclosed technology. In addition, it is not intended to be bound by any express or implied theory presented in the aforementioned technical field, background technology or the following specific embodiments.

[0027] In the following detailed description of the embodiments, many specific details are set forth in order to provide a more thorough understanding of the disclosed technology. However, it is apparent to one of ordinary skill in the art that the disclosed technology can be practiced without these specific details. In other instances, well-known features are not described in detail to avoid unnecessarily complicating the description.

[0028] Terms such as "comprising" and "including" indicate that in addition to the units and steps directly and clearly stated in the specification, the technical solution of the present application does not exclude the situation of having other units and steps that are not directly or clearly stated. Terms such as "first" and "second" do not indicate the order of units in terms of time, space, size, etc., but are only used to distinguish between units.

[0029] "User profile data" in this article refers to a collection of information used to describe and characterize user characteristics, preferences, behavior patterns, etc. This data is usually constructed by collecting, collating and analyzing multiple data sources of users. Exemplarily, user profile data may include one or more of the following: user identification, artificial statistical information (e.g., age, gender, etc.), interest preferences, social relationships, and historical playback data.

[0030] The "user-perceived information" herein refers to dynamic data that can reflect the user's current state. Exemplarily, the user-perceived information includes one or more of the following: the user's current location, time, weather, activity scene (e.g., sports, work, study, rest, party, etc.), social status (e.g., alone, with friends, family gathering, etc.), emotional state (e.g., happy, sad, calm, excited, etc.), function usage scenario (e.g., whether wearing headphones, driving, in a public place, etc.), calendar information (e.g., birthday, anniversary).

[0031] Natural Language Processing (NLP) is the study of computer systems that can effectively implement natural language communication, and is an important direction in the fields of computer science and artificial intelligence. Large Language Model (LLM) refers to a deep learning model trained with a large amount of text data, which can generate natural language text or understand the meaning of language text. Large language models can handle a variety of natural language tasks, such as text classification, question answering, dialogue, etc., and are an important path to artificial intelligence.

[0032] Hereinafter, various exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings.

[0033] Referring to the accompanying drawings, Figure 1 1 is a schematic flowchart of a music recommendation method 10 according to one or more embodiments of the present application.

[0034] It should be noted that the execution subject of the music recommendation method 10 may be a hardware device with data information processing capabilities and / or the necessary software required to drive the hardware device to work. Optionally, the execution subject may include a workstation, a server, a computer, a user terminal and other intelligent devices. Among them, the user terminal includes but is not limited to a mobile phone, a computer, an intelligent voice interaction device, an intelligent home appliance, a vehicle-mounted terminal, etc.

[0035] like Figure 1 As shown, in step S110, a candidate music list and user perception information of authorized access are obtained.

[0036] The candidate music list refers to a collection of candidate music that the user may be interested in, determined based on the user request and user profile data. Optionally, the candidate music list here not only includes basic information of the candidate music (e.g., name, singer, album, etc.), but may also be preliminarily sorted according to a specific algorithm.

[0037] In one possible implementation, the candidate music list is generated by the execution subject of method 10. Exemplarily, the client of the execution subject first receives a user request submitted by the user through an interface or other interactive methods. The request may be an input text or a voice message. In the case where the user request is a voice message, the voice is converted into text by automatic speech recognition (ASR). After that, the user request is parsed using an NLP algorithm to determine the user request semantics. Exemplarily, the user request semantics include on-demand requests and recommendation requests, where on-demand requests usually contain clear song or singer names, while recommendation requests are more general, such as "play some songs". Next, a recall algorithm that matches the user request semantics is determined. A recall algorithm is a method for quickly retrieving information related to a user request from a large amount of data. Exemplarily, for on-demand requests, a text matching algorithm (e.g., regular expressions, full-text search engines, etc.) can be used to search for music matching the user request in a music database; for recommendation requests, deep learning, collaborative filtering, and other algorithms can be used to select candidate music matching the user request semantics and user portrait data from the music database to generate a candidate music list.

[0038] In another possible implementation, the candidate music list is provided by a third-party music platform. Exemplarily, the execution subject of method 10 inputs the user request or its corresponding semantics into the third-party music platform through a search interface, and receives the music or candidate music list returned by the third-party music platform.

[0039] In step S120, the candidate music list, user perception information and user portrait data are converted into descriptive text that meets the requirements of natural language processing.

[0040] "Descriptive text that meets the requirements of natural language processing" refers to text that follows the rules of natural language and is easy to be understood and processed by the NLP system. It uses the language used by humans in daily communication (such as Chinese, English, etc.) to describe and express various information, rather than using encoding, marking or other non-natural language forms. For example, the song information in the candidate music list may be converted into "This song is sung by the famous singer Jay Chou, belongs to the pop music genre, and the lyrics depict a youthful love story." User perception information may be converted into "The current user is driving home and it is drizzling outside." User portrait data may be converted into "The user is a post-90s generation, has recently played cheerful songs, and has shared many tracks with positive emotions."

[0041] These descriptive texts that meet the requirements of natural language processing provide rich contextual information for the large language model, which helps to generate more personalized music recommendations that meet user expectations. In addition, this transformation achieves the unification of feature dimensions, making it simple to add new user features when the system is subsequently updated, because any new features can be converted into descriptive text and input into the model without changing the input layer structure of the model, and reducing the complexity of model training. Exemplarily, the transformation technologies involved in this step may include but are not limited to: natural language generation, template filling, conditional text generation, knowledge graphs and semantic analysis, or pre-trained language models.

[0042] In step S130, the descriptive text is embedded into the first prompt information and input into the first large language model to obtain a recommended music ranking, wherein the recommended music is a subset of the candidate music.

[0043] Prompt information, also known as prompt, is used to guide large language models (e.g., GPT series, BERT series, etc.) to understand and process descriptive text. It can be designed according to specific task requirements, and usually includes a description of the task, an explanation of the input data, etc. Since the candidate music list, user perception information, and user portrait data have been converted into descriptive text in step S120, in a possible implementation, the above descriptive text can be simply embedded into a preset prompt information template to generate the first prompt information.

[0044] As described above, user perception information includes multiple categories. For example, user perception information can be divided into location, time, weather, activity scene, social status, emotional state, function usage scenario, and calendar information. It is understandable that if all user perception information is added to the prompt, the model calculation amount will be very large. Therefore, a weight can be assigned to each type of user perception information, which helps to give priority to more important perception information in the screening process of candidate music. Optionally, the size of the weight may depend on the importance and / or probability of occurrence of the information. For example, since the probability of occurrence of the user perception information "Today is the user's anniversary" is much lower than that of "the user is driving", the weight of the former is greater than that of the latter. In addition, the more important the user perception information is, the higher the weight. Next, based on the determined weights, the descriptive text of each type of user perception information can be screened using a weighted sampling algorithm. It is understandable that if the user perception information is filtered only according to the weight, the categories of the user perception information filtered out each time are relatively similar, and the user perception information with low weight is easy to be ignored. Therefore, a weighted sampling algorithm can be used when performing category filtering, so as to enrich the diversity of recommendations as much as possible under the premise of being constrained by the weight. Exemplarily, the weighted sampling algorithm used in the embodiment of the present application can be a weighted random sampling or a reservoir sampling algorithm. Finally, the descriptive text of the filtered user perception information is respectively embedded into the preset prompt information template to generate the first prompt information.

[0045] The first language model is a deep learning model that has been trained and can handle music recommendation tasks, such as GPT-3, BERT, etc. Optionally, the first language model selects recommended music from the candidate music based on the degree of association between the candidate music in the candidate music list and the user perception information and user portrait data, and sorts the recommended music based on the degree of association to output the recommended music ranking. Since the model input information contains user perception information, the recommendation results are more in line with the user's current actual needs and personalized preferences. For example, the user is currently in Shanghai at night, then if the candidate music list contains the song "Night Shanghai", this song will have a high probability of appearing in the recommended music.

[0046] In addition, after obtaining the recommended music ranking, the recommended music can be finely arranged again according to the user perception information, so as to strengthen the degree of association between the user perception information and the music recommendation. It is understandable that the quality of the first song in the recommended music ranking directly affects the user's preliminary evaluation of the entire recommendation result. If the first song can attract the user's attention and leave a good impression, then the user will be more likely to continue to try other songs in the recommendation list. In this regard, the first song in the recommended music ranking can be determined based on the user perception information during the fine arrangement. In a possible implementation, the recommended music ranking output by the first language model is converted into a descriptive text that meets the requirements of natural language processing, and the descriptive text of the recommended music ranking and the user perception information is embedded in the second prompt information, and input into the second language model, so as to select the music with the highest degree of association with the user perception information from the recommended music, as the first place in the recommended music ranking. Optionally, in the process of embedding the descriptive text of the user perception information into the second prompt information, the weighted sampling algorithm described above can also be used to screen the descriptive texts of various types of user perception information, and the specific screening process can refer to the above, which will not be repeated here.

[0047] In step S140, the recommended music ranking is returned to the user.

[0048] Optionally, after the recommended music ranking is presented to the user, if the user does not request other music within a preset time, the music ranked first in the recommended music ranking is directly played.

[0049] Furthermore, in order to enhance the connection between the user and the music content and improve the user's perception of the music recommendation service, a large language model can be used to fuse music information, user perception information, etc. to generate reasons for recommending the music to the user (i.e., recommendation words), and the recommendation words are presented to the user during the prelude of each song, giving the user a more immersive listening experience. In one possible implementation, the music to be played is first determined, which may be the music that ranks first in the recommended music ranking or the music requested by the user. Next, the introduction information of the music to be played and the descriptive text of the user's perception information are embedded in the third prompt information and input into the third large language model to obtain the recommendation words for the music to be played, and the recommendation words are presented in an interactive manner. Optionally, in order to prevent the recommendation words from being generated according to a fixed pattern, multiple prompt information templates can be preset, each of which can have different sentence patterns and tones, and a prompt information template is randomly selected when generating the third prompt information. For example, the prompt information template may be "You are a music DJ with cutting-edge music recommendation skills, specializing in creating personalized music experiences. Your task is to generate a song introduction of no more than 50 words based on the song title, singer name and the following user perception information. The song introduction will make users feel relaxed and happy and enjoy the music better." Exemplarily, the introduction angle may also be incorporated into the third prompt information, and the introduction angle may be associated with the user portrait information. For example, if the user is a post-2000s generation and a star chaser, "Entertainment news about singers" may be added to the third prompt information as an introduction angle.

[0050] It should be noted that the first language model, the second language model and the third language model mentioned above can be different models trained by different training data, or they can be the same model that has the ability to generate recommended music rankings, the ability to fine-tune recommended music, and the ability to generate recommendation words.

[0051] In addition, when presenting the recommendation to the user, one or more of the image information, voice information, expression information and action information of the client's virtual image can be determined based on the recommendation of the music to be played, and the client can be controlled to interact based on one or more of the image information, voice information, expression information and action information. For example, the cartoon image, timbre, expression, displayed action, etc. of the vehicle-mounted robot interacting with the user can be controlled to give the vehicle-mounted robot vitality and improve the user experience.

[0052] Continue to refer Figure 2 , Figure 2 1 is a schematic flowchart of a music recommendation method 20 according to one or more embodiments of the present application.

[0053] In step S210, a candidate music list is obtained. The details of obtaining the candidate music list can be found in the description of S110 above, which will not be repeated here.

[0054] In step S220, music is sorted. When sorting music, the candidate music list, user perception information, and user portrait data may first be converted into descriptive text that meets the requirements of natural language processing, and then the descriptive text is embedded in the first prompt information and input into the first language model to obtain the recommended music sorting. The execution of music sorting may refer to the description of S120-S130 above, which will not be repeated here.

[0055] In step S230, music is finely arranged. The fine arrangement process is intended to strengthen the degree of association between user perception information and music recommendations, and the first song in the recommended music ranking can also be determined based on the user perception information during the fine arrangement. When performing the fine arrangement, the recommended music ranking output by the first language model can be converted into a descriptive text that meets the requirements of natural language processing, and the descriptive text of the recommended music ranking and the user perception information is embedded in the second prompt information and input into the second language model to select the music with the highest degree of association with the user perception information from the recommended music as the first in the recommended music ranking. For the execution of the music ranking, please refer to the description of S130 above, which will not be repeated here.

[0056] In step S240, a recommendation is generated. A recommendation can be generated for the music to be played (for example, the music ranked first in the recommended music ranking, or the music requested by the user), and the recommendation can be presented to the user during the prelude when playing, giving the user a more immersive listening experience. When generating the recommendation, the introduction information of the music to be played and the descriptive text of the user's perception information can be embedded in the third prompt information and input into the third language model to obtain the recommendation for the music to be played. For details about the steps for generating the recommendation, please refer to the above description of Figure 1 The description is omitted here. The music recommendation method 10, 20 according to one or more embodiments of the present application integrates user perception information into the consideration of screening recommended music, so that the recommendation results are more in line with the user's current actual needs and personalized preferences, thereby improving user satisfaction and overall experience, and avoiding the disadvantage of difficulty in accurately recommending new users during cold start. In addition, in the process of screening and sorting recommended music, the music recommendation method 10, 20 converts all input information into descriptive text that meets the requirements of natural language processing. This conversion makes it simple to add new user features, because any new features can be converted into descriptive text and input into the model without changing the input layer structure of the model, and reduces the complexity of model training.

[0057] Figure 33 is a schematic block diagram of a music recommendation device 30 according to one or more embodiments of the present application. The music recommendation device 30 includes a memory 310, a processor 320, and a computer program 330 stored in the memory 310 and executable on the processor 320. The execution of the computer program 330 enables the music recommendation method 10 or 20 to be executed. The music recommendation device 30 may be a workstation, computer, user terminal, or other intelligent device with data information processing capabilities. The user terminal includes but is not limited to a mobile phone, a computer, an intelligent voice interaction device, an intelligent home appliance, a vehicle-mounted terminal, and the like.

[0058] In addition, as described above, the present application can also be implemented as a computer storage medium, in which there is stored a program for causing a computer to execute the following steps: Figure 1 The music recommendation method 10 shown or Figure 2 Here, as the computer storage medium, various computer storage media such as disks (e.g., magnetic disks, optical disks, etc.), cards (e.g., memory cards, optical cards, etc.), semiconductor memories (e.g., ROMs, nonvolatile memories, etc.), and tapes (e.g., magnetic tapes, cassettes, etc.) can be used.

[0059] The present application may also be implemented as a computer program product, which includes a computer program. When the computer program is executed by a processor, the following Figure 1 The music recommendation method 10 shown or Figure 2 The procedure of the music recommendation method 20 is shown.

[0060] The relevant user personal information that may be involved in the various embodiments of this application is strictly in accordance with the requirements of laws and regulations, following the principles of legality, legitimacy and necessity, based on the reasonable purposes of business scenarios, to process the personal information that users actively provide during the use of products / services or generated due to the use of products / services, as well as the personal information obtained with the user's authorization.

[0061] The user personal information processed by this application will vary depending on the specific product / service scenario, and will be based on the specific scenario in which the user uses the product / service. It may involve the user's account information, device information, driving information, vehicle information or other related information. The applicant will treat the user's personal information and its processing with a high degree of diligence.

[0062] This application attaches great importance to the security of user personal information and has taken reasonable and feasible security protection measures that meet industry standards to protect user information and prevent personal information from being accessed, disclosed, used, modified, damaged or lost without authorization.

[0063] In the applicable situation, hardware, software or a combination of hardware and software can be used to realize the various embodiments provided by the application. Moreover, in the applicable situation, without departing from the scope of the application, the various hardware components and / or software components set forth herein can be combined into a composite component comprising software, hardware and / or both. In the applicable situation, without departing from the scope of the application, the various hardware components and / or software components set forth herein can be divided into subcomponents comprising software, hardware or both. In addition, in the applicable situation, it is contemplated that the software component can be implemented as a hardware component, and vice versa.

[0064] Software (such as program code and / or data) according to the present application can be stored on one or more computer storage media. It is also contemplated that the software identified herein can be implemented using one or more general or special-purpose computers and / or computer systems, networked and / or otherwise. Where applicable, the order of the various steps described herein can be changed, combined into composite steps, and / or divided into sub-steps to provide the features described herein.

[0065] The embodiments and examples set forth herein are provided to best illustrate embodiments according to the present application and its specific applications, and thereby enable those skilled in the art to implement and use the present application. However, those skilled in the art will appreciate that the above description and examples are provided only for ease of illustration and example. The description set forth is not intended to cover all aspects of the present application or to limit the present application to the precise form disclosed.

Claims

1. A music recommendation method, characterized in that: The method comprises the following steps: Acquire a candidate music list and user perception information authorized to access, wherein the candidate music list includes candidate music determined based on a user request and user portrait data; Converting the candidate music list, the user perception information and the user portrait data into descriptive text that meets the requirements of natural language processing; embedding the descriptive text into the first prompt information and inputting the result into the first large language model to obtain a recommended music ranking, wherein the recommended music is a subset of the candidate music; and The recommended music ranking is returned to the user.

2. The music recommendation method according to claim 1, characterized in that: The user perception information includes one or more of the following: user's current location, time, weather, activity scene, function usage scenario, and calendar information.

3. The music recommendation method according to claim 1, characterized in that: Embedding the descriptive text into the first prompt information and inputting it into the first language model to obtain the recommended music ranking includes: Selecting recommended music from the candidate music based on the degree of association between the candidate music and the user perception information and the user portrait data; and The recommended music is ranked based on the association degree to determine the recommended music ranking.

4. The music recommendation method according to claim 1, characterized in that: The method further comprises: Utilize natural language processing algorithms to parse user requests to determine the semantics of user requests; Determining a recall algorithm that semantically matches the user request; and The recall algorithm is used to select candidate music that matches the user request semantics and the user portrait data from a music database to generate the candidate music list.

5. The music recommendation method according to claim 1, characterized in that: The method further comprises: Converting the recommended music ranking into descriptive text that meets the requirements of natural language processing; embedding the recommended music ranking and the descriptive text of the user perception information into the second prompt information, and inputting the second largest language model to select the music with the highest correlation with the user perception information from the recommended music; and The music with the highest correlation degree is listed as the first in the recommended music ranking.

6. The music recommendation method according to claim 1 or 5, characterized in that: In the process of embedding the descriptive text of the user-perceived information into the first prompt information and / or the second prompt information, the following operations are performed: Determine the weight of each type of user perception information; Based on the determined weights, a weighted sampling algorithm is used to screen the descriptive texts of various types of user perception information; The descriptive text of the filtered user-perceived information is embedded in the first prompt information and / or the second prompt information.

7. The music recommendation method according to claim 1, characterized in that: The method further comprises: Determining music to be played, wherein the music to be played is the music ranked first in the recommended music ranking or the music requested by the user; embedding the introduction information of the music to be played and the descriptive text of the user perception information into the third prompt information, and inputting the third language model to obtain a recommendation for the music to be played; The recommendation is presented in an interactive manner.

8. The music recommendation method according to claim 7, characterized in that: The method further comprises: A prompt information template is randomly selected from a set of prompt information templates with different sentence patterns as the third prompt information.

9. The music recommendation method according to claim 7, characterized in that: The method further comprises: Determining one or more of image information, voice information, expression information, and action information of a virtual image of the client based on the recommendation words of the music to be played; and The client is controlled to interact based on one or more of the image information, the voice information, the expression information and the action information.

10. A music recommendation device, characterized in that: The invention comprises: a memory; a processor; and a computer program stored in the memory and executable on the processor, wherein the execution of the computer program enables the music recommendation method according to any one of claims 1 to 9 to be executed.

11. A computer storage medium, characterized in that: The computer storage medium includes instructions, which, when executed, execute the music recommendation method according to any one of claims 1-9.

12. A computer program product, characterized in that It comprises a computer program, which, when executed by a processor, implements the music recommendation method according to any one of claims 1-9.

Citation Information

Patent Citations

  • Music recommendation method based on voiceprint and context awareness and terminal equipment

    CN110083738A

  • Music recommendation method and device, electronic equipment and computer readable storage medium

    CN114357236A

  • Music recommendation method, recommendation system, intelligent cabin and vehicle

    CN116450946A

  • Content recommendation method and device based on voice interaction, equipment and storage medium

    CN116719988A

  • Vehicle-mounted music recommendation method and device and computer storage medium

    CN117076758A

Cited By

  • AI interaction method and device, wearable equipment and storage medium

    CN119920249A

  • AI Interaction Method and Apparatus, Wearable Device, Storage Medium

    CN119920249B