Model training method and device, equipment and storage medium

Abstract

The embodiment of the invention relates to a model training method and device, equipment and a storage medium. The method provided herein includes constructing a set of candidate lyric content based on reference lyric content, each candidate lyric content including at least one paragraph in the reference lyric content; based on the evaluation information of the group of candidate lyric contents, determining target lyric contents meeting a preset requirement from the group of candidate lyric contents; generating description information corresponding to the target lyric content, wherein the description information indicates multiple attributes of the target lyric content; based on the description information, constructing a group of cue words corresponding to the target lyric content; and training a lyric generation model based on the group of cue words and the target lyric content. In this way, according to the embodiment of the invention, the reference lyric content can be continuously screened and iterated through multiple dimensions and steps, the quality of the reference lyric content is improved, and the lyric generation effect is optimized.

Description

Technical Field

[0001] The exemplary embodiments disclosed herein relate generally to the field of computers, and more particularly to methods, apparatus, devices, and computer-readable storage media for training models. Background Technology

[0002] With the development of the internet and computer technology, natural language processing has also flourished. In the field of natural language processing, lyric generation models have received widespread attention and application. Therefore, the generation quality of these models has become a focal point of concern. Summary of the Invention

[0003] In a first aspect of this disclosure, a method for training a model is provided. The method includes: constructing a set of candidate lyrics based on reference lyrics, each candidate lyrics including at least one paragraph from the reference lyrics; determining target lyrics that meet preset requirements from the set of candidate lyrics based on evaluation information of the set of candidate lyrics; generating descriptive information corresponding to the target lyrics, the descriptive information indicating multiple attributes of the target lyrics; constructing a set of prompt words corresponding to the target lyrics based on the descriptive information; and training a lyrics generation model based on the set of prompt words and the target lyrics.

[0004] In a second aspect of this disclosure, an apparatus for training a model is provided. The apparatus includes: a first construction module configured to construct a set of candidate lyrics based on reference lyrics, each candidate lyric including at least one paragraph from the reference lyrics; a lyrics determination module configured to determine target lyrics that meet preset requirements from the set of candidate lyrics based on evaluation information of the set of candidate lyrics; an information generation module configured to generate descriptive information corresponding to the target lyric, the descriptive information indicating multiple attributes of the target lyric; a second construction module configured to construct a set of prompt words corresponding to the target lyric based on the descriptive information; and a model training module configured to train a lyrics generation model based on the set of prompt words and the target lyric.

[0005] In a third aspect of this disclosure, an electronic device is provided. The device includes at least one processing unit; and at least one memory coupled to the at least one processing unit and storing instructions for execution by the at least one processing unit. When executed by the at least one processing unit, the instructions cause the device to perform the method of the first aspect.

[0006] In a fourth aspect of this disclosure, a computer-readable storage medium is provided. The computer-readable storage medium stores a computer program that can be executed by a processor to implement the method of the first aspect.

[0007] It should be understood that the content described in this content section is not intended to limit the key or essential features of the embodiments of this disclosure, nor is it intended to restrict the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0008] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. In the drawings, the same or similar reference numerals denote the same or similar elements, wherein:

[0009] Figure 1 A schematic diagram is shown of an example environment in which embodiments of the present disclosure may be implemented;

[0010] Figure 2 A flowchart illustrating an example training model process according to some embodiments of the present disclosure is shown;

[0011] Figure 3 A schematic diagram of an example training model according to some embodiments of the present disclosure is shown;

[0012] Figure 4 A schematic structural block diagram of an example training model apparatus according to some embodiments of the present disclosure is shown; and

[0013] Figure 5 A block diagram of an electronic device capable of implementing several embodiments of the present disclosure is shown. Detailed Implementation

[0014] Embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this disclosure. It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure.

[0015] It should be noted that the headings of any section / subsection provided herein are not limiting. Various embodiments are described throughout this document, and embodiments of any type may be included under any section / subsection. Furthermore, embodiments described in any section / subsection may be combined in any way with any other embodiments described in the same section / subsection and / or different sections / subsections.

[0016] In the description of embodiments of this disclosure, the term "comprising" and similar terms should be understood as open-ended inclusion, i.e., "including but not limited to". The term "based on" should be understood as "at least partially based on". The term "one embodiment" or "the embodiment" should be understood as "at least one embodiment". The term "some embodiments" should be understood as "at least some embodiments". Other explicit and implicit definitions may also be included below. The terms "first", "second", etc., may refer to different or the same objects. Other explicit and implicit definitions may also be included below.

[0017] The embodiments of this disclosure may involve user data, data acquisition, and / or use. All of these aspects comply with applicable laws, regulations, and relevant provisions. In the embodiments of this disclosure, all data collection, acquisition, processing, manipulation, forwarding, and use are conducted with the user's knowledge and confirmation. Accordingly, in implementing the embodiments of this disclosure, the type, scope of use, and usage scenarios of any data or information that may be involved should be communicated to the user and their authorization obtained in accordance with relevant laws and regulations through appropriate means. The specific methods of notification and / or authorization may vary depending on the actual situation and application scenario, and the scope of this disclosure is not limited in this respect.

[0018] In this specification and the embodiments, any processing of personal information will be carried out only under the premise of legality (such as obtaining the consent of the personal information subject, or being necessary for the performance of a contract), and will only be carried out within the scope stipulated or agreed upon. A user's refusal to process personal information other than that necessary for basic functions will not affect the user's use of basic functions.

[0019] In this specification and the embodiments, if the training and inference of the model are involved, the data involved (including but not limited to the data itself, the acquisition and / or use of the data) comply with the requirements of relevant laws and regulations.

[0020] Traditional approaches have two main drawbacks. First, lyric generation models trained on standard text cannot meet the diverse requirements of lyrics. Second, training electronic devices solely on lyric data cannot achieve high adaptability to downstream song generation tasks.

[0021] Embodiments of this disclosure propose a training model scheme. According to this scheme, a set of candidate lyrics can be constructed based on reference lyrics, each candidate lyric including at least one paragraph from the reference lyrics; based on evaluation information of the set of candidate lyrics, target lyrics that meet preset requirements are determined from the set of candidate lyrics; descriptive information corresponding to the target lyric is generated, the descriptive information indicating multiple attributes of the target lyric; based on the descriptive information, a set of prompt words corresponding to the target lyric is constructed; and based on the set of prompt words and the target lyric, a lyrics generation model is trained.

[0022] In this way, embodiments of this disclosure can construct training data with multi-dimensional attributes (e.g., lyric theme, song style, vocal information, expression state, and lyric structure) based on reference lyrics content, thereby improving the quality of the training data. Furthermore, by using such training data to train the lyrics generation model, embodiments of this disclosure can improve the quality of the lyrics generated by the lyrics generation model and give it music-related multi-dimensional attributes, thereby improving its adaptability to music generation models.

[0023] The following section provides a detailed description of various example implementations of this scheme, with reference to the accompanying drawings.

[0024] Example Environment

[0025] Figure 1 A schematic diagram of an example environment 100 in which embodiments of the present disclosure can be implemented is shown. For example... Figure 1 As shown, the example environment 100 may include an electronic device 110 and a lyrics generation model 120.

[0026] In this example environment 100, electronic device 110 constructs training data based on reference lyrics to train lyrics generation model 120. Electronic device 110 is at least configured to construct a set of candidate lyrics from the received reference lyrics. Further, electronic device 110 determines target lyrics and a corresponding set of prompt words based on this set of candidate lyrics. Electronic device 110 trains lyrics generation model 120 based on the target lyrics and the set of prompt words.

[0027] As an example, such a lyrics generation model 120 could be a transformer-based language model.

[0028] In some embodiments, the electronic device 110 may establish a communication connection with the lyrics generation model 120. That is, the electronic device 110 may invoke the local or remote lyrics generation model 120.

[0029] In some embodiments, electronic device 110 may be any type of mobile terminal, fixed terminal, or portable terminal, including mobile phones, desktop computers, laptop computers, notebook computers, netbook computers, tablet computers, media computers, multimedia tablets, personal communication system (PCS) devices, personal navigation devices, personal digital assistants (PDAs), audio / video players, digital cameras / camcorders, positioning devices, television receivers, radio receivers, e-book devices, gaming devices, or any combination thereof, including accessories and peripherals of these devices or any combination thereof. In some embodiments, electronic device 110 may also support any type of user-facing interface (such as "wearable" circuitry).

[0030] It should be understood that the structure and function of the various elements in environment 100 are described for illustrative purposes only and do not imply any limitation on the scope of this disclosure.

[0031] Example process

[0032] Figure 2 A flowchart of an example process 200 for training a model according to some embodiments of the present disclosure is shown. Process 200 can be implemented at electronic device 110. Reference is made below. Figure 1 To describe process 200.

[0033] In some embodiments, the electronic device 110 can acquire reference music content. Furthermore, the electronic device 110 can identify the reference music content as corresponding reference audio content and reference lyrics content based on a pre-trained model.

[0034] like Figure 2 As shown in box 210, electronic device 110 constructs a set of candidate lyrics content based on reference lyrics content, each candidate lyrics content including at least one paragraph from the reference lyrics content.

[0035] In some embodiments, such reference lyrics can be lyrics in any language. Figure 3 In box 311, electronic device 110 can perform preprocessing operations on the reference lyrics content.

[0036] In some embodiments, the electronic device 110 may determine multiple paragraphs of the reference lyrics content. As an example, the electronic device 110 may utilize an audio processing model (e.g., a Deep Chorus model) to process the reference music content to classify the paragraph types of the reference lyrics content corresponding to the reference music content. As an example, such paragraph types may include verses, choruses, and other paragraph types.

[0037] In some embodiments, the electronic device 110 can utilize a pre-trained model (e.g., a tagging model) to identify the musical style and expressive state in the reference audio content. Such musical styles may include, for example, pop music, traditional Chinese music, rock music, etc. Furthermore, the electronic device 110 can also utilize a language model to infer vocal information in the reference audio content.

[0038] Furthermore, the electronic device 110 can construct multiple paragraph combinations of the multiple paragraphs obtained above to obtain a set of candidate lyrics content.

[0039] As an example, the electronic device 110 can retain the verse and chorus sections of the reference lyrics and randomly combine their corresponding multiple paragraphs to obtain a set of candidate lyrics. In some embodiments, such a set of candidate lyrics corresponds to a preset time length. Such a preset length can be, for example, 30 seconds to 90 seconds, to accommodate the need for generating 1 minute of lyrics.

[0040] In other embodiments, the electronic device 110 may not divide the reference lyrics into paragraphs, but directly use the reference lyrics (e.g., the complete lyrics) as candidate lyrics.

[0041] Reference Figure 2 In box 220, electronic device 110 determines target lyrics content that meets preset requirements from a set of candidate lyrics content based on evaluation information of a set of candidate lyrics content.

[0042] For ease of description, this set of candidate lyrics will be referred to as the first set of lyrics below. In some embodiments, such evaluation information may be, for example, text repetition information, rhyme evaluation information, and text fluency information.

[0043] In some embodiments, reference Figure 3 In box 312, electronic device 110 can use a basic text feature extraction model to filter out candidate lyrics that do not meet expectations or ideal conditions in the first group of candidate lyrics, so as to optimize the overall data quality.

[0044] Furthermore, the electronic device 110 can also remove at least one candidate lyric from the first group of candidate lyric content based on the evaluation information to determine the second group of candidate lyric content.

[0045] In some embodiments, the electronic device 110 may remove first candidate lyrics content with a text repetition rate higher than a first threshold from the first group of candidate lyrics content based on the text repetition rate indicated by the evaluation information. As an example, the electronic device 110 may filter data with excessively high text repetition in the first group of candidate lyrics content to remove first candidate lyrics content with a text repetition rate higher than the first threshold, thereby avoiding repetitive paradigms that are overfitted by the model.

[0046] In other embodiments, the electronic device 110 can remove second candidate lyrics with a rhyme score below a second threshold from the first group of candidate lyrics based on the rhyme score indicated by the evaluation information. As an example, the electronic device 110 can extract rhyme features (e.g., the pinyin of the last character of a line and rhyme scheme) from the first group of candidate lyrics and calculate the rhyme score of the lyric segment. Further, the electronic device 110 can remove second candidate lyrics with a rhyme score below the second threshold, retaining candidate lyrics with a clear rhyming effect.

[0047] In some other embodiments, the electronic device 110 can remove third candidate lyrics content whose text fluency is lower than a third threshold from the first group of candidate lyrics content based on the text fluency indicated by the evaluation information. As an example, the electronic device 110 can extract text fluency features from the first group of candidate lyrics content and use a language model to calculate perplexity as a text fluency index. Further, the electronic device 110 can remove third candidate lyrics content whose text fluency index is lower than the third threshold, retaining candidate lyrics content whose fluency meets preset requirements.

[0048] In this way, the electronic device 110 can determine the second set of candidate lyrics based on the text repetition, rhyme evaluation, and the relationship between text fluency and threshold in the first set of candidate lyrics.

[0049] In some embodiments, the electronic device 110 can determine the target lyrics content from the second set of candidate lyrics content obtained above.

[0050] refer to Figure 3 In box 313, electronic device 110 can generate a topic description text for the second set of candidate lyrics. As an example, electronic device 110 can use a language model to score the harmlessness features of the second set of candidate lyrics, removing candidate lyrics containing harmful content. Furthermore, electronic device 110 can also use a pre-trained topic generation model to generate the general meaning of the lyrics after removing harmful content from the second set of candidate lyrics, thus obtaining the topic description text for the second set of candidate lyrics.

[0051] In some embodiments, the lyrics summary generation stage requires generating a summary for each candidate lyric in the second group of candidate lyrics to simulate possible user input in real-world scenarios. In some embodiments, the electronic device 110 can utilize a language model to generate topic data for each candidate lyric through prompting engineering. In some embodiments, each candidate lyric generates a "story-based" topic (a short story with more text) and a "keyword" topic (a short keyword). In this way, the electronic device 110 can cover a wide range of topic input formats. Furthermore, the electronic device 110 can perform fine-tuning of this data to ensure data quality, and train a language model based on this high-quality data to generate topic data for the candidate lyric content. The electronic device 110 can also add control conditions on the input side to achieve controllable generation of "story-based" or "keyword" topics. Even further, the electronic device 110 can use the trained language model to generate "story-based" and "keyword" topics from a large number of candidate lyric contents, and mix the two topics in a certain proportion as the final lyrics summary, which serves as the topic description text for the second group of candidate lyric contents.

[0052] Furthermore, the electronic device 110 can determine the target lyrics content based on the degree of matching between the topic description text and the reference lyrics content. As an example, the electronic device 110 can score the relevance between the generated topic description text and the reference lyrics content based on a language model, removing data with low topic relevance, thereby determining the target lyrics content.

[0053] Furthermore, the electronic device 110 can re-score the confidence of the musical style and the confidence of the expression state of the target lyrics based on the language model, and reassign the musical style and expression state labels based on the scoring results, thereby improving the overall relevance of the musical style and expression state of the target lyrics and the coverage of niche types. In this way, the electronic device 110 can achieve calibration and equalization of the musical style and expression state of the target lyrics.

[0054] Reference Figure 2 In box 230, electronic device 110 generates descriptive information corresponding to the target lyrics content, the descriptive information indicating multiple attributes of the target lyrics content.

[0055] In some embodiments, such multiple attributes may include, for example, lyrical theme, song style, vocal information, expressive state, and lyrical structure.

[0056] In some embodiments, the electronic device 110 may provide the first model with reference descriptions of preset attributes of the target lyrics content to generate a set of extended descriptions corresponding to the preset attributes.

[0057] In some embodiments, such preset attributes may be, for example, song style, expression state, or lyric structure.

[0058] refer to Figure 3 In box 314, electronic device 110 can provide the first model with reference descriptions of the song style of the target lyrics content to expand the song style of the target lyrics content, increase the mapping ability from any style to a set of styles, and enhance the diversity of song styles.

[0059] In some embodiments, the electronic device 110 may also provide the first model with reference descriptions of the expression state of the target lyrics content, so as to expand the expression state of the target lyrics content, increase the mapping ability from any expression state to a set of expression states, and enhance the diversity of expression states.

[0060] In some embodiments, the electronic device 110 may also provide the first model with a reference description of the lyrics structure of the target lyrics content to expand the lyrics structure of the target lyrics content, increase the structure control capability, and enrich the structure control situation.

[0061] In some embodiments, the lyrics structure control diversity enhancement step simulates the structure control instructions under real user input conditions and trains the first model to follow the instructions or default strategies under different conditions. In some embodiments, the electronic device 110 designs default strategies for different real instruction situations, including situations where both structure and line number are specified simultaneously, situations where only line number is specified, situations where only structure is specified, or situations where neither is specified. In some embodiments, the electronic device 110 can also randomly simulate real instruction situations according to the default strategy to ensure that the training data is sufficiently diverse and can cover as many real situations as possible. In some embodiments, the electronic device 110 can randomly add structures that need to be ignored, such as intro tracks and accompaniment, to train the first model's ability to ignore such structures.

[0062] In summary, the electronic device 110 can obtain a set of extended descriptions of the target lyrics based on the reference descriptions corresponding to the first model, the song style, expression state, and lyric structure of the target lyrics.

[0063] In this way, the electronic device 110 can generate description information corresponding to the target lyrics content based on the reference description content and the set of extended description content obtained above.

[0064] Reference Figure 2 In box 240, electronic device 110 constructs a set of prompt words corresponding to the target lyrics content based on the description information.

[0065] In some embodiments, the descriptive information indicates multiple attributes of the target lyrics content. (See reference) Figure 3 In box 315, electronic device 110 can construct multiple attribute combinations based on the aforementioned multiple attributes. As an example, electronic device 110 can randomly mask one or more attributes from the multiple attributes, including lyric theme, song style, vocal information, expression state, and lyric structure, to obtain multiple attribute combinations. In this way, electronic device 110 can simulate real user input to train the model's ability to automatically complete missing input information.

[0066] In some embodiments, the electronic device 110 may provide the combination of these multiple attributes to the second model to generate a set of prompt words.

[0067] In box 315, electronic device 110 can use a language model to refine the set of prompt words obtained above and generate a natural language description to simulate the user's real input.

[0068] Reference Figure 2 In box 250, electronic device 110 trains a lyrics generation model based on a set of prompt words and target lyrics content.

[0069] In some embodiments, the electronic device 110 can design the output format of the lyrics generation model. As an example, the electronic device 110 can concatenate the lyrics theme, song style, expression state, vocal information, lyrics structure, and target lyrics content (e.g., a thought chain format) as the output format of the lyrics generation model.

[0070] Continue to refer to Figure 3 In box 316, electronic device 110 can supervise and fine-tune the lyrics generation model based on a set of prompts and target lyrics content.

[0071] In summary, the electronic device 110 can expand the input part from basic features to arbitrary forms of natural language instructions based on the training method of the thought chain. The output part first extracts features and then generates lyrics, thereby getting rid of the dependence on the front-end module and realizing an end-to-end lyrics generation model.

[0072] In this way, embodiments of this disclosure can construct training data with multi-dimensional attributes based on reference lyrics content, thereby improving the training quality of the lyrics generation model. Embodiments of this disclosure can generate lyrics content based on the trained lyrics generation model. Furthermore, electronic devices can extract or infer key music-related attributes as input to the music generation model, thereby improving the adaptability of the music generation model.

[0073] Example devices and equipment

[0074] Embodiments of this disclosure also provide corresponding apparatus for implementing the above methods or processes. Figure 4 A schematic structural block diagram of an example training model apparatus 400 according to certain embodiments of the present disclosure is shown. Apparatus 400 may be implemented as or included in electronic device 110. Various modules / components in apparatus 400 may be implemented by hardware, software, firmware, or any combination thereof.

[0075] like Figure 4 As shown, the device 400 includes a first construction module 410 configured to construct a set of candidate lyrics based on reference lyrics, each candidate lyrics including at least one paragraph from the reference lyrics; a lyrics determination module 420 configured to determine target lyrics that meet preset requirements from the set of candidate lyrics based on evaluation information of the set of candidate lyrics; an information generation module 430 configured to generate descriptive information corresponding to the target lyrics, the descriptive information indicating multiple attributes of the target lyrics; a second construction module 440 configured to construct a set of prompt words corresponding to the target lyrics based on the descriptive information; and a model training module 450 configured to train a lyrics generation model based on the set of prompt words and the target lyrics.

[0076] In some embodiments, the first construction module 410 is further configured to determine multiple paragraphs of reference lyrics content; and to construct multiple paragraph combinations to obtain a set of candidate lyrics content.

[0077] In some embodiments, a set of candidate lyrics is a first set of lyrics, and the lyrics determination module 420 is further configured to remove at least one candidate lyric from the first set of candidate lyrics based on evaluation information to determine a second set of candidate lyrics; and to determine the target lyrics from the second set of candidate lyrics.

[0078] In some embodiments, the lyrics determination module 420 is further configured to perform at least one of the following: removing first candidate lyrics content with a text repetition higher than a first threshold from the first group of candidate lyrics content based on the text repetition indicated by the evaluation information; removing second candidate lyrics content with a rhyme evaluation lower than a second threshold from the first group of candidate lyrics content based on the rhyme evaluation indicated by the evaluation information; and removing third candidate lyrics content with a text fluency lower than a third threshold from the first group of candidate lyrics content based on the text fluency indicated by the evaluation information.

[0079] In some embodiments, the lyrics determination module 420 is further configured to generate a second set of topic description texts for candidate lyrics content; and to determine target lyrics content based on the degree of matching between the topic description texts and reference lyrics content.

[0080] In some embodiments, the information generation module 430 is further configured to provide the first model with reference description content about preset attributes of the target lyrics content, so as to generate a set of extended description content corresponding to the preset attributes; and to generate description information corresponding to the target lyrics content based on the reference description content and the set of extended description content.

[0081] In some embodiments, the multiple attributes of the target lyrics content indicated by the description information include multiples of the following: lyric theme, song style, vocal information, expression state, and lyric structure.

[0082] In some embodiments, the second construction module 440 is further configured to construct multiple attribute combinations of multiple attributes; and to generate a set of prompt words based on the multiple attribute combinations.

[0083] In some embodiments, the second building module 440 is also configured to provide the second model with a combination of attributes to generate a set of prompt words.

[0084] In some embodiments, a set of candidate lyrics corresponds to a preset time length.

[0085] The modules included in device 400 can be implemented in various ways, including software, hardware, firmware, or any combination thereof. In some embodiments, one or more units can be implemented using software and / or firmware, such as machine-executable instructions stored on a storage medium. In addition to or as an alternative to machine-executable instructions, some or all of the modules in device 400 can be implemented at least partially by one or more hardware logic components. By way of example, and not limitation, exemplary types of hardware logic components that can be used include field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), system-on-a-chip (SoCs), complex programmable logic devices (CPLDs), and so on.

[0086] Figure 5 A block diagram of an electronic device 500 in which one or more embodiments of the present disclosure may be implemented is shown. It should be understood that... Figure 5 The electronic device 500 shown is merely exemplary and should not be construed as limiting the functionality and scope of the embodiments described herein. Figure 5 The electronic device 500 shown can be used to achieve Figure 1 Electronic devices 110.

[0087] like Figure 5As shown, electronic device 500 is in the form of a general-purpose electronic device. Components of electronic device 500 may include, but are not limited to, one or more processors or processing units 510, memory 520, storage device 530, one or more communication units 540, one or more input devices 550, and one or more output devices 560. Processing unit 510 may be a physical or virtual processor and is capable of performing various processes according to programs stored in memory 520. In a multiprocessor system, multiple processing units execute computer-executable instructions in parallel to improve the parallel processing capability of electronic device 500.

[0088] Electronic device 500 typically includes multiple computer storage media. Such media can be any accessible media that is accessible to electronic device 500, including but not limited to volatile and non-volatile media, removable and non-removable media. Memory 520 can be volatile memory (e.g., registers, cache, random access memory (RAM)), non-volatile memory (e.g., read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory), or some combination thereof. Storage device 530 can be removable or non-removable media and can include machine-readable media, such as flash drives, disks, or any other media that can be used to store information and / or data and can be accessed within electronic device 500.

[0089] Electronic device 500 may further include additional removable / non-removable, volatile / non-volatile storage media. Although not explicitly stated... Figure 5 As shown, disk drives for reading from or writing to removable, non-volatile disks (e.g., "floppy disks") and optical disk drives for reading from or writing to removable, non-volatile optical disks can be provided. In these cases, each drive can be connected to a bus (not shown) via one or more data media interfaces. Memory 520 may include computer program product 525 having one or more program modules configured to perform various methods or actions of various embodiments of this disclosure.

[0090] Communication unit 540 enables communication with other electronic devices via a communication medium. Additionally, the functionality of components of electronic device 500 can be implemented using a single computing cluster or multiple computing machines capable of communicating via communication connections. Therefore, electronic device 500 can operate in a networked environment using logical connections to one or more other servers, network personal computers (PCs), or another network node.

[0091] Input device 550 can be one or more input devices, such as a mouse, keyboard, trackball, etc. Output device 560 can be one or more output devices, such as a monitor, speaker, printer, etc. Electronic device 500 can also communicate with one or more external devices (not shown) via communication unit 540 as needed. These external devices include storage devices, display devices, etc., and can communicate with one or more devices that enable user interaction with electronic device 500, or with any device that enables electronic device 500 to communicate with one or more other electronic devices (e.g., network card, modem, etc.). Such communication can be performed via input / output (I / O) interface (not shown).

[0092] According to an exemplary implementation of this disclosure, a computer-readable storage medium is provided that stores computer-executable instructions thereon, wherein the computer-executable instructions are executed by a processor to implement the methods described above. According to an exemplary implementation of this disclosure, a computer program product is also provided, which is tangibly stored on a non-transitory computer-readable medium and includes computer-executable instructions, which are executed by a processor to implement the methods described above.

[0093] Various aspects of this disclosure are described herein with reference to flowchart illustrations and / or block diagrams of methods, apparatuses, devices, and computer program products implemented according to this disclosure. It should 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-readable program instructions.

[0094] These computer-readable program instructions can be provided to a processing unit of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that, when executed by the processing unit of the computer or other programmable data processing apparatus, they create means for implementing the functions / actions specified in one or more blocks of the flowchart and / or block diagram. These computer-readable program instructions can also be stored in a computer-readable storage medium that causes a computer, programmable data processing apparatus, and / or other device to operate in a particular manner. Thus, the computer-readable medium storing the instructions comprises an article of manufacture that includes instructions for implementing aspects of the functions / actions specified in one or more blocks of the flowchart and / or block diagram.

[0095] Computer-readable program instructions can be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions that execute on the computer, other programmable data processing apparatus, or other device to perform the functions / actions specified in one or more boxes of a flowchart and / or block diagram.

[0096] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of an instruction, which contains one or more executable instructions for implementing the specified logical function. In some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0097] Various implementations of this disclosure have been described above. The foregoing description is exemplary and not exhaustive, nor is it limited to the disclosed implementations. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described implementations. The terminology used herein is chosen to best explain the principles, practical applications, or improvements to technology in the market, or to enable others skilled in the art to understand the various implementations disclosed herein.

Claims

1. A method for training a model, comprising: constructing a set of candidate lyrics content based on a reference lyrics content, each candidate lyrics content comprising at least one passage in the reference lyrics content; determining a target lyrics content satisfying a preset requirement from the set of candidate lyrics content based on evaluation information of the set of candidate lyrics content; generating description information corresponding to the target lyrics content, the description information indicating a plurality of attributes of the target lyrics content; constructing a set of prompt words corresponding to the target lyrics content based on the description information; and training a lyrics generation model based on the set of prompt words and the target lyrics content.

2. The method of claim 1, wherein the constructing a set of candidate lyrics content based on a reference lyrics content comprises: determining a plurality of passages of the reference lyrics content; and constructing a plurality of passage combinations of the plurality of passages to obtain the set of candidate lyrics content.

3. The method of claim 1, wherein the set of candidate lyrics content is a first set of lyrics content, and the determining a target lyrics content satisfying a preset requirement from the set of candidate lyrics content based on evaluation information of the set of candidate lyrics content comprises: removing at least one candidate lyrics content from the first set of candidate lyrics content based on the evaluation information to determine a second set of candidate lyrics content; and determining the target lyrics content from the second set of candidate lyrics content.

4. The method of claim 3, wherein the removing at least one candidate lyrics content from the first set of candidate lyrics content based on the evaluation information to determine a second set of candidate lyrics content comprises at least one of: removing a first candidate lyrics content having a text repetition degree higher than a first threshold from the first set of candidate lyrics content based on the text repetition degree indicated by the evaluation information; removing a second candidate lyrics content having a rhyme evaluation lower than a second threshold from the first set of candidate lyrics content based on the rhyme evaluation indicated by the evaluation information; removing a third candidate lyrics content having a text fluency degree lower than a third threshold from the first set of candidate lyrics content based on the text fluency degree indicated by the evaluation information.

5. The method of claim 3, wherein the determining the target lyrics content from the second set of candidate lyrics content comprises: generating a theme description text of the second set of candidate lyrics content; and determining the target lyrics content based on a matching degree of the theme description text and the reference lyrics content.

6. The method of claim 1, wherein the generating description information corresponding to the target lyrics content further comprises: providing a reference description content about a preset attribute of the target lyrics content to a first model to generate a set of extended description content corresponding to the preset attribute; and generating the description information corresponding to the target lyrics content based on the reference description content and the set of extended description content.

7. The method of claim 1, wherein the plurality of attributes of the target lyrics content indicated by the description information comprises a plurality of the following: ​ ​ ​ ​ ​ lyric theme, song style, vocal information, expression state, lyric structure. 8.The method of claim 1, wherein constructing a set of prompt words corresponding to the target lyric content based on the description information comprises: constructing a plurality of attribute combinations of the plurality of attributes; and generating the set of prompt words based on the plurality of attribute combinations. 9.The method of claim 8, wherein generating the set of prompt words based on the plurality of attribute combinations comprises: providing the plurality of attribute combinations to a second model to generate the set of prompt words. 10.The method of claim 1, wherein the set of candidate lyric contents corresponds to a preset time length. 11.An apparatus for training a model, comprising: a first constructing module configured to construct a set of candidate lyric contents based on a reference lyric content, each candidate lyric content including at least one passage in the reference lyric content; a lyric determining module configured to determine a target lyric content satisfying a preset requirement from the set of candidate lyric contents based on evaluation information of the set of candidate lyric contents; an information generating module configured to generate description information corresponding to the target lyric content, the description information indicating a plurality of attributes of the target lyric content; a second constructing module configured to construct a set of prompt words corresponding to the target lyric content based on the description information; and a model training module configured to train a lyric generation model based on the set of prompt words and the target lyric content. 12.An electronic device, comprising: at least one processing unit; and at least one memory coupled to the at least one processing unit and storing instructions for execution by the at least one processing unit, the instructions, when executed by the at least one processing unit, cause the electronic device to perform the method according to any one of claims 1 to 10. 13.A computer-readable storage medium having stored thereon a computer program, the computer program being executable by a processor to implement the method according to any one of claims 1 to 10.

Images