A song recommendation method, system, device, and medium

Abstract

This application discloses a song recommendation method, system, device, and medium, relating to the field of data processing technology. The method includes: acquiring historical song preference information and playback record information of the current listening playlist of a target user, wherein the playback record information includes: song metadata and user behavior information corresponding to each song; converting the song metadata and user behavior information of multiple songs into structured context text in chronological order; fusing the context text with the historical song preference information to obtain prompt words for a large language model; and using the prompt words as input to the large language model so that the large language model outputs song recommendation results for the target user.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a song recommendation method, system, device, and medium. Background Technology

[0002] Currently, with the rapid development of digital music platforms and the exponential growth of music libraries, users often struggle to efficiently discover songs that match their personal preferences when faced with a vast amount of music content. To address this challenge, mainstream song recommendation systems typically employ traditional recommendation algorithms such as collaborative filtering, content-based feature matching, or user historical behavior statistics. These methods analyze users' long-term listening records, favorites lists, or rating data to build user interest models, thereby predicting and recommending music content that users may be interested in.

[0003] However, existing song recommendation methods struggle to deeply understand users' real-time listening intentions and contextual situations. In particular, when users are listening to a specific playlist or are in a particular emotional state, traditional methods often fail to provide nuanced and timely recommendations, thus affecting the accuracy of recommendations and user satisfaction. Summary of the Invention

[0004] The summary section introduces a series of simplified concepts, which will be further explained in detail in the detailed description section. This summary section is not intended to limit the key features and essential technical features of the claimed technical solution, nor is it intended to determine the scope of protection of the claimed technical solution.

[0005] In a first aspect, embodiments of this application provide a song recommendation method, the method comprising: Obtain the target user's historical song preference information and the playback record information of the current listening playlist. The playback record information includes: song metadata and user behavior information for each song. Transform the song metadata and user behavior information of multiple songs into structured contextual text in chronological order; The contextual text is fused with the historical song preference information to obtain the prompt words of the large language model; The prompt words are used as input to the large language model so that the large language model outputs song recommendation results for the target user.

[0006] In one embodiment of the present invention, the step of using the prompt word as input to the large language model, so that the large language model outputs song recommendation results for the target user, includes: The prompt words are input into a large language model, which then performs semantic understanding and intent reasoning based on the prompt words, style continuity requirements, emotional connection requirements, and user behavior pattern response requirements. The large language model then outputs song recommendation results for the target user. The style continuity requirement refers to whether the next song continues the style of the current listening playlist. The emotional connection requirement is to predict the target user's emotional direction based on the playback history information. The user behavior pattern response requirement is for the target user to switch songs or add songs to their favorites in the current listening playlist.

[0007] In one embodiment of the present invention, the song recommendation result includes a list of recommended songs and the recommendation reasons for each song in the list.

[0008] In one embodiment of the present invention, after using the prompt words as input to the large language model to enable the large language model to output song recommendation results for the target user, the process includes: The recommended song results are inserted into the playback queue of the current listening playlist and displayed to the target user.

[0009] In one embodiment of the present invention, after inserting the recommended song results into the playback queue of the current listening playlist and displaying them to the target user, the process includes: Obtain feedback behavior data from the target user regarding the recommended song results, including playback completion rate or collection action.

[0010] In one embodiment of the present invention, after obtaining the feedback behavior data of the target user on the recommended song results, the process includes: The large language model is trained based on the feedback behavior data.

[0011] In one embodiment of the present invention, the song metadata includes at least one of song title, artist, album, and style tag, and the user behavior information includes at least one of playback completion, favorite operation, share operation, song skipping operation, and playback time point.

[0012] Secondly, this application proposes a song recommendation system, which includes: a data acquisition module, a data fusion module, and a song recommendation module; The data acquisition module is configured to acquire the target user's historical song preference information and the playback record information of the current listening playlist. The playback record information includes: song metadata and user behavior information corresponding to each song. The data fusion module is configured to: convert the song metadata and user behavior information of multiple songs into structured context text in chronological order; and fuse the context text with the historical song preference information to obtain prompt words for the large language model. The song recommendation module is configured to take the prompt words as input to the large language model, so that the large language model outputs song recommendation results for the target user.

[0013] Thirdly, a vehicle includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program stored in the memory to implement the steps of a song recommendation method as described in any of the first aspects above.

[0014] Fourthly, this application also proposes a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, it implements the steps of a song recommendation method according to any one of the first aspects.

[0015] In summary, the song recommendation method of this application obtains the target user's historical song preference information and the playback record information of the current listening playlist. The playback record information includes the song meta-information and user behavior information corresponding to each song. Then, the song meta-information and user behavior information of multiple songs are transformed into structured contextual text in chronological order. This can accurately capture the target user's emotional trend when listening to the current playlist. After fusing the contextual text and historical song preference information, the data is input into a large language model. The powerful deep semantic understanding and reasoning capabilities of the large language model are fully utilized to analyze the target user's true intention when listening to the current playlist. Finally, the method outputs song recommendation results that highly match the target user's mood, thereby improving the accuracy of song recommendation results and user satisfaction.

[0016] The song recommendation method proposed in this application, along with other advantages, objectives, and features of this application, will be partly apparent from the following description and partly understood by those skilled in the art through study and practice of this application. Attached Figure Description

[0017] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit this specification. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 This is a flowchart illustrating a song recommendation method provided in an embodiment of this application. Figure 2This application provides a schematic diagram of the structure of a song recommendation system. Figure 3 This is a schematic diagram of the structure of a song recommendation electronic device provided in an embodiment of this application. Detailed Implementation

[0018] To better understand the technical solutions provided in the embodiments of this specification, the technical solutions of the embodiments of this specification will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments of this specification and the specific features in the embodiments are detailed descriptions of the technical solutions of the embodiments of this specification, rather than limitations on the technical solutions of this specification. In the absence of conflict, the embodiments of this specification and the technical features in the embodiments can be combined with each other.

[0019] In this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, without necessarily requiring or implying any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element. The term "two or more" includes two or more cases.

[0020] Please see Figure 1 This is a flowchart illustrating a song recommendation method provided in an embodiment of this application, specifically including: S110. Obtain the target user's historical song preference information and the playback record information of the current listening playlist. The playback record information includes: song metadata and user behavior information corresponding to each song.

[0021] For example, it's necessary to simultaneously collect the target user's historical song preferences and the playback history of their current playlist. The playback history includes metadata for each song and user behavior related to those songs. In a car music recommendation scenario, if a user has a long-standing preference for jazz and blues and frequently listens to artist x, this represents their historical song preferences. The playback history of the current playlist includes: song A played in its entirety and saved; song B played 85% and then skipped without being saved; and song C played in its entirety without being saved. This approach considers both the user's long-term music preferences and their current listening behavior, avoiding recommendations that only align with the past or focus solely on the present, thus improving the accuracy of song recommendations.

[0022] S120. Transform the song metadata and user behavior information of multiple songs into structured context text in chronological order.

[0023] For example, after obtaining the playback history information of the target user's current listening playlist, it is necessary to convert the song metadata and user behavior information of the multiple songs contained therein into a structured contextual text according to the actual playback time order of the songs in the playlist. The essence of this conversion process is to reconstruct the song metadata and user behavior information, such as song title, artist, style tag, playback completion, whether it is saved, and whether it is skipped, into a natural language description with temporal logic and semantic coherence. For example, for the playback history in the aforementioned example, the system would organize it as: The user is listening to the laid-back jazz playlist and has played in the following order: Song A (jazz, artist X, played in full, saved), Song B (cool jazz, artist Y, skipped after playing 85%, not saved), Song C (Bossa Nova, artist Z, played in full, not saved). Through this transformation method, the user's immediate attitude towards each song, implicit in the song's metadata and user behavior information (such as dislike expressed through skipping songs or liking expressed through adding songs to favorites), and the stylistic evolution of songs are completely converted into text form. This constructs a contextual text that accurately depicts the user's current music listening situation. Consequently, the large language model can accurately identify the target user's true intention when listening to the current playlist.

[0024] S130. The context text is fused with the historical song preference information to obtain the prompt words of the large language model.

[0025] For example, after generating contextual text that accurately describes the user's current music listening context, the system fuses this contextual text with pre-acquired historical song preference information of the target user. This historical song preference information represents a stable, long-term music interest profile formed by the user, including but not limited to features such as the user's long-term preferred artists, music styles, and frequently listened-to playlists. The essence of the fusion process is to structurally combine the dynamic contextual text describing the user's real-time listening behavior with the static historical profile representing the user's inherent preferences, constructing a large language model input prompt that simultaneously includes both the user's long-term interest dimension and the real-time context dimension. For instance, the contextual text describing the playback sequence of songs A, B, and C is concatenated or embedded with the user's historical preference for jazz and blues styles and frequently listened-to artist x to generate a complete input prompt. This allows the input of the large language model to simultaneously include both real-time listening context and long-term preference dimensions, making the basic information for the large language model's reasoning more comprehensive, thereby improving the relevance of the recommendation results.

[0026] S140. The prompt word is used as input to the large language model so that the large language model outputs song recommendation results for the target user.

[0027] For example, the fused prompts are input into a large language model, which then performs inference and outputs song recommendations tailored to the target user. In the aforementioned in-vehicle scenario, the fused prompts are input into a large language model fine-tuned for the music domain. This large language model includes either an LLaMA (Large Language Model Meta AI) model or a GPT (Generative Pre-trained Transformer) model. The large language model infers based on the information and outputs suitable jazz-style song recommendations. Leveraging the deep semantic understanding capabilities of large language models, the deep, non-linear relationship between user preferences and music content can be captured, addressing the limited representational capabilities of traditional recommendation models.

[0028] In summary, the song recommendation method proposed in this application obtains the target user's historical song preference information and the playback record information of the current listening playlist. The playback record information includes the song meta-information and user behavior information corresponding to each song. Then, the song meta-information and user behavior information of multiple songs are transformed into structured contextual text in chronological order. This can accurately capture the target user's emotional trend when listening to the current playlist. After fusing the contextual text and historical song preference information, the data is input into a large language model. The powerful deep semantic understanding and reasoning capabilities of the large language model are fully utilized to analyze the target user's true intention when listening to the current playlist. Finally, the method outputs song recommendation results that highly match the target user's mood, thereby improving the accuracy of song recommendation results and user satisfaction.

[0029] In some examples, using the prompt words as input to the large language model so that the large language model outputs song recommendation results for the target user includes: The prompt words are input into a large language model, which then performs semantic understanding and intent reasoning based on the prompt words, style continuity requirements, emotional connection requirements, and user behavior pattern response requirements. The large language model then outputs song recommendation results for the target user. The style continuity requirement refers to whether the next song continues the style of the current listening playlist. The emotional connection requirement is to predict the target user's emotional direction based on the playback history information. The user behavior pattern response requirement is for the target user to switch songs or add songs to their favorites in the current listening playlist.

[0030] For example, after inputting the fused prompts into a large language model fine-tuned with music domain data, the large language model first performs deep analysis of the prompts, identifying information from both long-term historical preferences and real-time listening context. Based on this, the large language model performs semantic understanding and intent inference according to three preset inference dimensions: style continuity requirement, emotional connection requirement, and user behavior pattern response requirement. Specifically, the style continuity requirement indicates that the large language model needs to determine whether the next recommended song should continue the overall style of the current playlist or introduce appropriate changes to maintain freshness; the emotional connection requirement requires the large language model to predict the emotional direction the target user might want to hear next based on the emotional characteristics of songs in the playback history, such as through song meta-information and the emotional direction implied by user behavior; the user behavior pattern response requirement means that the large language model needs to respond instantly to the user's specific actions on already played songs, such as skipping songs to express dissatisfaction or adding them to favorites to express liking, adjusting the recommendation strategy to match the user's instantaneous preference changes. The large language model integrates the inference results from these three dimensions and selects candidate songs that highly match the current context from a vast music knowledge base as the recommended song.

[0031] It can accurately capture the user's current listening intent and contextual characteristics, so that the song recommendation results not only match the overall style of the current playlist, but also respond to the user's instantaneous behavioral changes, solving the problem that traditional recommendation methods cannot perceive the user's real-time listening intent; at the same time, through the comprehensive judgment of three reasoning dimensions, the song recommendation results maintain stylistic consistency while also having a moderate sense of novelty, enhancing the user's listening immersion.

[0032] In some examples, the song recommendation results include a list of recommended songs and the reasons for recommending each song in the list.

[0033] For example, the song recommendation result includes a list of recommended songs and the recommendation reasons for each song in the list. Specifically, after inputting the fused prompt words into a large language model fine-tuned with music domain data, the large language model first performs deep semantic analysis on the prompt words, extracting the target user's historical song preference information and the user intent features composed of contextual text converted from the playback history information of the current listening playlist. Based on this, the large language model performs inference based on its music semantic understanding and knowledge association capabilities learned from large-scale music corpora and user interaction data. The large language model identifies the style trend, emotional context, and the user's immediate attitude towards each song in the listening playlist based on the song meta-information of the played songs in the context text and their corresponding user behavior information. Moreover, the large language model combines historical song preference information to improve the accuracy of the song recommendation results. Subsequently, the large language model selects several songs from the candidate song library that highly match the current context to generate a recommended song list. At the same time, the large language model also generates a natural language recommendation reason for each recommended song in the recommended song list, such as recommending this song because it has a similar lazy atmosphere to the song B that was just played, and it is a new work by artist X that you have saved.

[0034] It solves the problem of traditional recommendation systems that only recommend songs without explaining the reasons, allowing users to clearly understand the logic behind the recommendations and enhancing user satisfaction with the results. In in-vehicle scenarios, concise reasons for recommendations can be displayed through the vehicle's infotainment screen or voice broadcast, improving the user's interactive experience.

[0035] In some examples, after taking the prompt words as input to the large language model so that the large language model outputs song recommendations for the target user, the process includes: The recommended song results are inserted into the playback queue of the current listening playlist and displayed to the target user.

[0036] For example, after the large language model outputs song recommendation results for the target user, the recommended song results are inserted into the playback queue of the current listening playlist. Specifically, the recommended song results output by the large language model are parsed to obtain the song identifiers and corresponding playback metadata of each recommended song. Based on the playback progress and queue structure of the current listening playlist, the songs in the recommended song results are added to the subsequent playback positions of the current playlist according to a preset insertion method. The preset insertion method can be a sequential append mode, where the recommended song results are placed after the existing songs to be played in the current playlist; or an instant insertion mode, where the recommended song results are inserted as high-priority content after the currently playing song and before the next existing song to be played. Through the above insertion operation, the recommended song results are seamlessly connected with the user's currently listening songs, ensuring that the recommended content can be naturally integrated into the subsequent playback content without interrupting the user's listening experience.

[0037] In some examples, after inserting the recommended song results into the playback queue of the current listening playlist and displaying them to the target user, the process includes: Obtain feedback behavior data from the target user regarding the recommended song results, including playback completion rate or collection action.

[0038] For example, after inserting the recommended song into the playback queue of the current listening playlist and displaying it to the target user, the system monitors and collects the target user's feedback behavior data in real time. This feedback behavior data includes playback completion rate or save / favorite operation. Specifically, the process of collecting feedback behavior data aims to capture the user's immediate reaction to the recommended content. Playback completion rate represents the user's acceptance of the recommended song, i.e., whether the user played the entire recommended song or skipped it midway. Save / favorite operation represents the user's liking of the recommended song, i.e., whether the user triggered the save / favorite function during or after playback.

[0039] By collecting feedback behavior data from target users, the system can accurately determine users' preferences for recommended songs, solving the problem that traditional recommendation methods struggle to obtain real-time feedback and evaluate recommendation effectiveness. The collected feedback behavior data provides a real basis for target user behavior in subsequent model optimization, enabling recommended song results to accurately capture users' instantaneous changes in preferences.

[0040] In some examples, after obtaining the target user's feedback behavior data on the recommended song results, the process includes: The large language model is trained based on the feedback behavior data.

[0041] For example, after obtaining the target user's feedback behavior data on the recommended song results, a large language model needs to be trained based on this feedback behavior data. Specifically, the feedback behavior data includes playback completion rate and favorite operations, where playback completion rate quantifies the user's acceptance of the recommended song; that is, a user playing the song completely indicates positive acceptance, while skipping the song midway indicates negative feedback. The favorite operation directly represents the user's strong preference. Training the large language model based on the feedback behavior data makes the model's reasoning logic more closely aligned with the user's actual preferences, continuously improving the accuracy of the recommended song results.

[0042] In some examples, the song metadata includes at least one of song title, artist, album, and style tag, and the user behavior information includes at least one of playback completion, favorite operation, share operation, song skipping operation, and playback time.

[0043] For example, by acquiring the target user's historical song preference information and the playback history information of the current listening playlist, the target user's listening preferences and instantaneous intentions can be accurately reflected. The historical song preference information is used to characterize the user's long-term stable music interest profile, including but not limited to long-term preferred artists, music styles, and frequently listened-to playlists. The playback history information of the current listening playlist includes the song metadata corresponding to each song in the sequence of songs played in the playlist, as well as user behavior information. The song metadata specifically includes at least one of song title, artist, album, and style tags, while the user behavior information specifically includes at least one of playback completion rate, favorite actions, sharing actions, song skipping actions, and playback time.

[0044] like Figure 2 As shown, this application proposes a song recommendation system, which includes: a data acquisition module 21, a data fusion module 22, and a song recommendation module 23; The data acquisition module 21 is configured to acquire the target user's historical song preference information and the playback record information of the current listening playlist. The playback record information includes: the song metadata and user behavior information corresponding to each song. The data fusion module 22 is configured to: convert the song metadata and user behavior information of multiple songs into structured context text in chronological order; and fuse the context text with the historical song preference information to obtain prompt words for the large language model. The song recommendation module 23 is configured to use the prompt words as input to the large language model, so that the large language model outputs song recommendation results for the target user.

[0045] The effects of applying the aforementioned method in the above system can be found in the description of the aforementioned method embodiments, and will not be repeated here.

[0046] like Figure 3 As shown, this application embodiment also provides an electronic device 300, including a memory 310, a processor 320, and a computer program 311 stored in the memory 310 and executable on the processor. When the processor 320 executes the computer program 311, it implements the steps of any of the above-mentioned song recommendation methods.

[0047] Since the electronic device described in this embodiment is the device used to implement a song recommendation device in the embodiments of this application, those skilled in the art can understand the specific implementation method and various variations of the electronic device in this embodiment based on the method described in the embodiments of this application. Therefore, how the electronic device implements the method in the embodiments of this application will not be described in detail here. Any device used by those skilled in the art to implement the method in the embodiments of this application falls within the scope of protection of this application.

[0048] In practical implementation, when the computer program 311 is executed by the processor, it can achieve the following: Figure 1 Any of the corresponding implementation methods in the embodiments.

[0049] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0050] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-readable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-readable program code.

[0051] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create a machine for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1A device that provides the functions specified in one or more boxes.

[0052] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0053] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0054] This application also provides a computer program product, which includes computer software instructions that, when executed on a processing device, cause the processing device to execute the LDPC decoding method of a solid-state drive controller.

[0055] A computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the flow or function according to the embodiments of this application is generated. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions may be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can store or a data storage device such as a server or data center that integrates one or more available media. The available medium may be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state disk (SSD)).

[0056] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0057] In the several embodiments provided in this application, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, or indirect coupling or communication connection between devices or units, and may be electrical, mechanical, or other forms.

[0058] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0059] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0060] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0061] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

[0062] Although preferred embodiments have been described in this specification, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this specification.

[0063] Obviously, those skilled in the art can make various modifications and variations to this specification without departing from its spirit and scope. Therefore, if such modifications and variations fall within the scope of the claims and their equivalents, this specification is also intended to include such modifications and variations.

Claims

1. A song recommendation method, characterized in that, The method includes: Obtain the target user's historical song preference information and the playback record information of the current listening playlist. The playback record information includes: song metadata and user behavior information for each song. Transform the song metadata and user behavior information of multiple songs into structured contextual text in chronological order; The contextual text is fused with the historical song preference information to obtain the prompt words of the large language model; The prompt words are used as input to the large language model so that the large language model outputs song recommendation results for the target user.

2. The song recommendation method according to claim 1, characterized in that, The step of using the prompt words as input to the large language model, so that the large language model outputs song recommendation results for the target user, includes: The prompt words are input into a large language model, which then performs semantic understanding and intent reasoning based on the prompt words, style continuity requirements, emotional connection requirements, and user behavior pattern response requirements. The large language model then outputs song recommendation results for the target user. The style continuity requirement refers to whether the next song continues the style of the current listening playlist. The emotional connection requirement is to predict the target user's emotional direction based on the playback history information. The user behavior pattern response requirement is for the target user to switch songs or add songs to their favorites in the current listening playlist.

3. The song recommendation method according to claim 1, characterized in that, The song recommendation results include a list of recommended songs and the reasons for recommending each song in the list.

4. The song recommendation method according to claim 1, characterized in that, After taking the prompt words as input to the large language model so that the large language model outputs song recommendation results for the target user, the process includes: The recommended song results are inserted into the playback queue of the current listening playlist and displayed to the target user.

5. The song recommendation method according to claim 4, characterized in that, After inserting the recommended song results into the playback queue of the current listening playlist and displaying them to the target user, the process includes: Obtain feedback behavior data from the target user regarding the recommended song results, including playback completion rate or collection action.

6. The song recommendation method according to claim 5, characterized in that, After obtaining the target user's feedback behavior data on the recommended song results, the process includes: The large language model is trained based on the feedback behavior data.

7. The song recommendation method according to claim 1, characterized in that, The song metadata includes at least one of song title, artist, album, and style tags, and the user behavior information includes at least one of playback completion, favorite operation, share operation, song skipping operation, and playback time.

8. A song recommendation system, characterized in that, The system includes: a data acquisition module, a data fusion module, and a song recommendation module; The data acquisition module is configured to acquire the target user's historical song preference information and the playback record information of the current listening playlist. The playback record information includes: song metadata and user behavior information corresponding to each song. The data fusion module is configured to: convert the song metadata and user behavior information of multiple songs into structured context text in chronological order; and fuse the context text with the historical song preference information to obtain prompt words for the large language model. The song recommendation module is configured to take the prompt words as input to the large language model, so that the large language model outputs song recommendation results for the target user.

9. A vehicle comprising: The memory and processor are characterized in that the processor is used to implement the steps of a song recommendation method as described in any one of claims 1-7 when executing a computer program stored in the memory.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of a song recommendation method as described in any one of claims 1-7.

Images