Lyric information processing method and device, storage medium and electronic equipment

Abstract

The invention relates to a lyric information processing method and device, a storage medium and electronic equipment, and relates to the technical field of big data processing, and the method comprises the steps: matching original complete lyric information corresponding to coded lyric information from a preset song database; according to a coded lyric line in the coded lyric information, matching an uncoded lyric line corresponding to the coded lyric line from the original complete lyric information; checking the accuracy of the code printing content in the coded lyric line according to the lyric line which is not printed with the code; and when it is determined that the code printing content of the coded lyric line has an error, recovering the code printing content of the coded lyric line according to the lyric line which is not coded. According to the invention, the recovery efficiency of the coding content is improved.

Description

Technical Field

[0001] The embodiments of the present disclosure relate to the technical field of big data processing. More specifically, the embodiments of the present disclosure relate to a method for processing lyrics information, a device for processing lyrics information, a computer-readable storage medium, and an electronic device. Background Art

[0002] This section is intended to provide a background or context to the embodiments of the disclosure that are recited in the claims. No description herein is admitted to be prior art by inclusion in this section.

[0003] In the existing method for recovering lyrics information, professional staff can manually recover the coded lyrics. However, this method has the following defects: on the one hand, the efficiency of recovering lyrics information is low; on the other hand, the accuracy of the recovered lyrics is low. Summary of the invention

[0004] However, in the relevant technical solutions, on the one hand, the manual recovery of coded lyrics has the problem of low recovery efficiency; on the other hand, the recovery of coded lyrics based on artificial intelligence content generation has the problem of low accuracy of lyrics recovery results.

[0005] To this end, an improved method for processing lyric information is urgently needed, so as to match the original complete lyric information corresponding to the coded lyric information from a preset song database; then, based on the coded lyric lines in the coded lyric information, match the uncoded lyric lines corresponding to the coded lyric lines from the original complete lyric information; then, based on the uncoded lyric lines, verify the accuracy of the coded content in the coded lyric lines; finally, when it is determined that there are errors in the coded content of the coded lyric lines, restore the coded content of the coded lyric lines based on the uncoded lyric lines, thereby improving the accuracy of the recovery results on the basis of improving the recovery efficiency of the coded content.

[0006] In this context, embodiments of the present disclosure are intended to provide a method for processing lyrics information, a device for processing lyrics information, a computer-readable storage medium, and an electronic device.

[0007] According to one aspect of the present disclosure, a method for processing lyrics information is provided, comprising:

[0008] Matching the original complete lyrics information corresponding to the coded lyrics information from a preset song database;

[0009] According to the coded lyrics line in the coded lyrics information, matching the uncoded lyrics line corresponding to the coded lyrics line from the original complete lyrics information;

[0010] Verifying the accuracy of the coded content in the coded lyrics line according to the uncoded lyrics line;

[0011] When it is determined that the coded content of the coded lyrics line has errors, the coded content of the coded lyrics line is restored according to the uncoded lyrics line.

[0012] In an exemplary embodiment of the present disclosure, matching the original complete lyrics information corresponding to the coded lyrics information from a preset song database includes:

[0013] Obtaining coded lyrics information of a current song and current song attribute information of the current song; wherein the current song attribute information includes at least one of the current song title, the current song singer, and the song album name to which the current song belongs;

[0014] According to the current song attribute information, candidate lyrics information corresponding to the current song is matched from a preset song database to obtain a first matching result, and the original complete lyrics information corresponding to the coded lyrics information is determined according to the first matching result.

[0015] In an exemplary embodiment of the present disclosure, determining the original complete lyrics information corresponding to the coded lyrics information according to the first matching result includes:

[0016] When it is determined that the first matching result is not empty, calculating a first text similarity between the coded lyrics information and the candidate lyrics information, and determining the original complete lyrics information from the candidate lyrics information according to the first text similarity;

[0017] When it is determined that the first matching result is empty, matching the associated songs and lyrics context corresponding to the current song from the preset song database according to the current song attribute information to obtain a second matching result;

[0018] If the second matching result is not empty, the original complete lyrics information is determined according to the associated song and the lyrics context; if the second matching result is empty, the original complete lyrics information is generated according to a preset content generation model.

[0019] In an exemplary embodiment of the present disclosure, calculating the first text similarity between the coded lyrics information and the candidate lyrics information includes:

[0020] Calculating a first text similarity between the coded lyrics information and the candidate lyrics information based on a preset similarity algorithm; and / or

[0021] Calculating a first text similarity between the coded lyrics information and the candidate lyrics information based on a preset edit distance algorithm; and / or

[0022] The first text similarity between the coded lyrics information and the candidate lyrics information is calculated based on a preset similarity calculation model.

[0023] In an exemplary embodiment of the present disclosure, calculating the first text similarity between the coded lyrics information and the candidate lyrics information based on a preset similarity algorithm includes:

[0024] Performing word segmentation processing on the coded lyrics information to obtain a first word segmentation result, and constructing a first word segmentation set based on the first word segmentation result;

[0025] Performing word segmentation processing on the candidate lyrics information to obtain a second word segmentation result, and constructing a second word segmentation set based on the second word segmentation result;

[0026] An intersection word set and a union word set between the first word segmentation set and the second word segmentation set are calculated, and a first text similarity between the coded lyric information and the candidate lyric information is determined based on the intersection word set and the union word set.

[0027] In an exemplary embodiment of the present disclosure, the preset similarity calculation model includes a first embedding mapping layer, a similarity comparison layer, a row weight matching layer, and a similarity calculation layer;

[0028] The step of calculating the first text similarity between the coded lyrics information and the candidate lyrics information based on a preset similarity calculation model includes:

[0029] Calculate, based on the first embedding mapping layer, a first semantic text vector for each line of lyrics in the coded lyrics information and a second semantic text vector for each line of lyrics in the candidate lyrics information;

[0030] Calculate the similarity between the first semantic text vector and the second semantic text vector based on the similarity comparison layer to obtain line text similarity;

[0031] Matching a row weight value for the first semantic text vector based on the row weight matching layer according to the first row position of the lyrics corresponding to the first semantic text vector in the coded lyrics information;

[0032] Based on the similarity calculation layer, a weighted sum is performed on the line text similarity and the line weight value corresponding to the line text similarity to obtain the first text similarity.

[0033] In an exemplary embodiment of the present disclosure, the preset content generation large model includes a second embedding mapping layer, a coding layer, and a hybrid expert model layer;

[0034] The method of generating the original complete lyrics information according to the preset content generation model includes:

[0035] Performing text preprocessing on the coded lyrics information to obtain standard lyrics information, and generating lyrics context information according to preset content parameter prompt information corresponding to the coded lyrics information and current song attribute information;

[0036] Based on the second embedding mapping layer, the standard lyrics information is subjected to word embedding processing to obtain lyrics text features, and the lyrics context information is subjected to embedding mapping processing to obtain a context marker sequence, and based on the encoding layer, the lyrics text features and the context marker sequence are subjected to encoding processing to obtain a context overall representation;

[0037] Predicting the context marker sequence and the overall context representation based on the hybrid expert model layer to obtain a text prediction result, and matching the text prediction result in a first preset sensitive word library to obtain a third matching result;

[0038] If the third matching result is that the text prediction result exists in the first preset sensitive word library, the original complete lyrics information is generated according to the text prediction result.

[0039] In an exemplary embodiment of the present disclosure, according to the coded lyrics line in the coded lyrics information, matching the uncoded lyrics line corresponding to the coded lyrics line from the original complete lyrics information includes:

[0040] Determine the first line position of the coded lyrics line in the coded lyrics information, and match the uncoded lyrics line corresponding to the coded lyrics line from the original complete lyrics information based on the first line position.

[0041] In an exemplary embodiment of the present disclosure, verifying the accuracy of the coded content in the coded lyrics line according to the uncoded lyrics line includes:

[0042] Comparing the uncoded lyrics line with the coded lyrics line to obtain the coded content in the coded lyrics line, and inputting the uncoded lyrics line into a preset coded content prediction model to obtain a coding prediction result;

[0043] Verifying the accuracy of the coding prediction result, and when determining that the coding prediction result is accurate, comparing the coding prediction result with the coding content in the same line of lyrics as the coding prediction result to verify the accuracy of the coding content;

[0044] Among them, if the coding prediction result and the coding content in the same lyrics line as the coding prediction result are consistent, then the accuracy of the coding content of the coded lyrics line is determined to be correct; if the coding prediction result and the coding content in the same lyrics line as the coding prediction result are inconsistent, then it is determined that there is an error in the accuracy of the coding content of the coded lyrics line.

[0045] In an exemplary embodiment of the present disclosure, the accuracy of the coding prediction result is verified, including:

[0046] The accuracy of the coding prediction results is verified based on the second line position of each coding prediction result in the uncoded lyrics and the remaining lyrics in the uncoded lyrics except the coding prediction results.

[0047] In an exemplary embodiment of the present disclosure, restoring the coded content of the coded lyrics line according to the uncoded lyrics line includes:

[0048] Identify the coded content of the coded lyrics line with errors to obtain a content identification result, and obtain the lyrics to be replaced from the uncoded lyrics line according to the third line position of the coded lyrics line to which the content identification result belongs in the coded lyrics information and the identification position of the content identification result in the coded lyrics line;

[0049] The content identification result is replaced based on the lyrics to be replaced to restore the coded content of the coded lyrics line.

[0050] In an exemplary embodiment of the present disclosure, the method for processing lyrics information further includes:

[0051] Input the coded content and the uncoded lyrics corresponding to the coded content into a preset risk control detection model to obtain a sensitive content detection result, and perform a secondary risk control detection on the coded content according to the sensitive content detection result;

[0052] If the secondary risk control detection result is that the coded content is safe content, a content recovery request is generated according to the restored coded content, and the content recovery request is pushed to the client where the operation and maintenance personnel are located to restore the coded content online.

[0053] In an exemplary embodiment of the present disclosure, the method for processing lyrics information further includes:

[0054] If the secondary risk control detection result is that the coded content is sensitive content, the uncoded lyrics line is segmented to obtain a second segmentation result, and the second segmentation result and the uncoded lyrics line are input into a preset sensitive word prediction model to obtain a first sensitive content detection result;

[0055] Input the word segmentation result and the uncoded lyrics into a second preset sensitive word library to obtain a second sensitive content detection result, and input the word segmentation result and the uncoded lyrics into a third preset sensitive word library to obtain a third sensitive content detection result;

[0056] Determining a target sensitive word from the word segmentation results according to the first sensitive content detection result, the second sensitive content detection result, and the third sensitive content detection result, and judging whether the target sensitive word is consistent with the coded content;

[0057] If the target sensitive word is consistent with the coded content, the coded content in the coded lyrics line is kept in a coded state; otherwise, a content recovery request is generated according to the restored coded content, and the content recovery request is pushed to the client where the operation and maintenance personnel are located to restore the coded content online.

[0058] According to one aspect of the present disclosure, a device for processing lyrics information is provided, comprising:

[0059] A complete lyrics information matching module, used to match the original complete lyrics information corresponding to the coded lyrics information from a preset song database;

[0060] An uncoded lyrics line matching module is used to match the uncoded lyrics line corresponding to the coded lyrics line from the original complete lyrics information according to the coded lyrics line in the coded lyrics information;

[0061] An accuracy verification module, used to verify the accuracy of the coded content in the coded lyrics line according to the uncoded lyrics line;

[0062] The coded content recovery module is used to recover the coded content of the coded lyrics line according to the uncoded lyrics line when it is determined that the coded content of the coded lyrics line has errors.

[0063] In an exemplary embodiment of the present disclosure, matching the original complete lyrics information corresponding to the coded lyrics information from a preset song database includes:

[0064] Obtaining coded lyrics information of a current song and current song attribute information of the current song; wherein the current song attribute information includes at least one of the current song title, the current song singer, and the song album name to which the current song belongs;

[0065] According to the current song attribute information, candidate lyrics information corresponding to the current song is matched from a preset song database to obtain a first matching result, and the original complete lyrics information corresponding to the coded lyrics information is determined according to the first matching result.

[0066] In an exemplary embodiment of the present disclosure, determining the original complete lyrics information corresponding to the coded lyrics information according to the first matching result includes:

[0067] When it is determined that the first matching result is not empty, calculating a first text similarity between the coded lyrics information and the candidate lyrics information, and determining the original complete lyrics information from the candidate lyrics information according to the first text similarity;

[0068] When it is determined that the first matching result is empty, matching the associated songs and lyrics context corresponding to the current song from the preset song database according to the current song attribute information to obtain a second matching result;

[0069] If the second matching result is not empty, the original complete lyrics information is determined according to the associated song and the lyrics context; if the second matching result is empty, the original complete lyrics information is generated according to a preset content generation model.

[0070] In an exemplary embodiment of the present disclosure, calculating the first text similarity between the coded lyrics information and the candidate lyrics information includes:

[0071] Calculating a first text similarity between the coded lyrics information and the candidate lyrics information based on a preset similarity algorithm; and / or

[0072] Calculating a first text similarity between the coded lyrics information and the candidate lyrics information based on a preset edit distance algorithm; and / or

[0073] The first text similarity between the coded lyrics information and the candidate lyrics information is calculated based on a preset similarity calculation model.

[0074] In an exemplary embodiment of the present disclosure, calculating the first text similarity between the coded lyrics information and the candidate lyrics information based on a preset similarity algorithm includes:

[0075] Performing word segmentation processing on the coded lyrics information to obtain a first word segmentation result, and constructing a first word segmentation set based on the first word segmentation result;

[0076] Performing word segmentation processing on the candidate lyrics information to obtain a second word segmentation result, and constructing a second word segmentation set based on the second word segmentation result;

[0077] An intersection word set and a union word set between the first word segmentation set and the second word segmentation set are calculated, and a first text similarity between the coded lyric information and the candidate lyric information is determined based on the intersection word set and the union word set.

[0078] In an exemplary embodiment of the present disclosure, the preset similarity calculation model includes a first embedding mapping layer, a similarity comparison layer, a row weight matching layer, and a similarity calculation layer;

[0079] The step of calculating the first text similarity between the coded lyrics information and the candidate lyrics information based on a preset similarity calculation model includes:

[0080] Calculate, based on the first embedding mapping layer, a first semantic text vector for each line of lyrics in the coded lyrics information and a second semantic text vector for each line of lyrics in the candidate lyrics information;

[0081] Calculate the similarity between the first semantic text vector and the second semantic text vector based on the similarity comparison layer to obtain line text similarity;

[0082] Matching a row weight value for the first semantic text vector based on the row weight matching layer according to the first row position of the lyrics corresponding to the first semantic text vector in the coded lyrics information;

[0083] Based on the similarity calculation layer, a weighted sum is performed on the line text similarity and the line weight value corresponding to the line text similarity to obtain the first text similarity.

[0084] In an exemplary embodiment of the present disclosure, the preset content generation large model includes a second embedding mapping layer, a coding layer, and a hybrid expert model layer;

[0085] The method of generating the original complete lyrics information according to the preset content generation model includes:

[0086] Performing text preprocessing on the coded lyrics information to obtain standard lyrics information, and generating lyrics context information according to preset content parameter prompt information corresponding to the coded lyrics information and current song attribute information;

[0087] Based on the second embedding mapping layer, the standard lyrics information is subjected to word embedding processing to obtain lyrics text features, and the lyrics context information is subjected to embedding mapping processing to obtain a context marker sequence, and based on the encoding layer, the lyrics text features and the context marker sequence are subjected to encoding processing to obtain a context overall representation;

[0088] Predicting the context marker sequence and the overall context representation based on the hybrid expert model layer to obtain a text prediction result, and matching the text prediction result in a first preset sensitive word library to obtain a third matching result;

[0089] If the third matching result is that the text prediction result exists in the first preset sensitive word library, the original complete lyrics information is generated according to the text prediction result.

[0090] In an exemplary embodiment of the present disclosure, according to the coded lyrics line in the coded lyrics information, matching the uncoded lyrics line corresponding to the coded lyrics line from the original complete lyrics information includes:

[0091] Determine the first line position of the coded lyrics line in the coded lyrics information, and match the uncoded lyrics line corresponding to the coded lyrics line from the original complete lyrics information based on the first line position.

[0092] In an exemplary embodiment of the present disclosure, verifying the accuracy of the coded content in the coded lyrics line according to the uncoded lyrics line includes:

[0093] Comparing the uncoded lyrics line with the coded lyrics line to obtain the coded content in the coded lyrics line, and inputting the uncoded lyrics line into a preset coded content prediction model to obtain a coding prediction result;

[0094] Verifying the accuracy of the coding prediction result, and when determining that the coding prediction result is accurate, comparing the coding prediction result with the coding content in the same line of lyrics as the coding prediction result to verify the accuracy of the coding content;

[0095] Among them, if the coding prediction result and the coding content in the same lyrics line as the coding prediction result are consistent, then the accuracy of the coding content of the coded lyrics line is determined to be correct; if the coding prediction result and the coding content in the same lyrics line as the coding prediction result are inconsistent, then it is determined that there is an error in the accuracy of the coding content of the coded lyrics line.

[0096] In an exemplary embodiment of the present disclosure, the accuracy of the coding prediction result is verified, including:

[0097] The accuracy of the coding prediction results is verified based on the second line position of each coding prediction result in the uncoded lyrics and the remaining lyrics in the uncoded lyrics except the coding prediction results.

[0098] In an exemplary embodiment of the present disclosure, restoring the coded content of the coded lyrics line according to the uncoded lyrics line includes:

[0099] Identify the coded content of the coded lyrics line with errors to obtain a content identification result, and obtain the lyrics to be replaced from the uncoded lyrics line according to the third line position of the coded lyrics line to which the content identification result belongs in the coded lyrics information and the identification position of the content identification result in the coded lyrics line;

[0100] The content identification result is replaced based on the lyrics to be replaced to restore the coded content of the coded lyrics line.

[0101] In an exemplary embodiment of the present disclosure, the lyric information processing device further includes:

[0102] A secondary risk control detection module is used to input the coded content and the uncoded lyrics corresponding to the coded content into a preset risk control detection model to obtain a sensitive content detection result, and perform a secondary risk control detection on the coded content according to the sensitive content detection result;

[0103] The first content recovery request generating module is used to generate a content recovery request according to the recovered coded content if the secondary risk control detection result is that the coded content is safe content, and push the content recovery request to the client where the operation and maintenance personnel are located, so as to restore the coded content online.

[0104] In an exemplary embodiment of the present disclosure, the lyric information processing device further includes:

[0105] A first sensitive content detection module, configured to, if the secondary risk control detection result is that the coded content is sensitive content, perform word segmentation processing on the uncoded lyrics line to obtain a second word segmentation result, and input the second word segmentation result and the uncoded lyrics line into a preset sensitive word prediction model to obtain a first sensitive content detection result;

[0106] A second sensitive content detection module, configured to input the word segmentation result and the uncoded lyrics into a second preset sensitive word library to obtain a second sensitive content detection result, and input the word segmentation result and the uncoded lyrics into a third preset sensitive word library to obtain a third sensitive content detection result;

[0107] a target sensitive word determination module, configured to determine a target sensitive word from the word segmentation result according to the first sensitive content detection result, the second sensitive content detection result, and the third sensitive content detection result, and to determine whether the target sensitive word is consistent with the coded content;

[0108] The second content recovery request generation module is used to keep the coded content in the coded lyrics line in a coded state if the target sensitive word is consistent with the coded content; otherwise, generate a content recovery request based on the restored coded content, and push the content recovery request to the client where the operation and maintenance personnel are located to restore the coded content online.

[0109] According to one aspect of the present disclosure, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the method for processing lyric information described in any one of the above is implemented.

[0110] According to one aspect of the present disclosure, there is provided an electronic device, including:

[0111] Processor; and

[0112] A memory, configured to store executable instructions of the processor;

[0113] The processor is configured to execute any one of the above-mentioned methods for processing lyrics information by executing the executable instructions.

[0114] According to the method for processing lyric information and the device for processing lyric information of the embodiment of the present disclosure, the original complete lyric information corresponding to the coded lyric information can be matched from a preset song database; then, based on the coded lyric lines in the coded lyric information, the uncoded lyric lines corresponding to the coded lyric lines are matched from the original complete lyric information; then, the accuracy of the coded content in the coded lyric lines is verified based on the uncoded lyric lines; finally, when it is determined that there is an error in the coded content of the coded lyric line, the coded content of the coded lyric line is restored based on the uncoded lyric line, without the need to manually restore the coded lyrics, thereby significantly reducing the problem of low recovery efficiency caused by manually restoring the coded lyrics, and reducing the problem of low accuracy of the recovery result caused by restoring the coded lyrics through artificial intelligence generated content, bringing a better experience to users. BRIEF DESCRIPTION OF THE DRAWINGS

[0115] The above and other objects, features and advantages of the exemplary embodiments of the present disclosure will become readily understood by reading the detailed description below with reference to the accompanying drawings. In the accompanying drawings, several embodiments of the present disclosure are shown in an exemplary and non-limiting manner, in which:

[0116] Figure 1 A flowchart of a method for processing lyrics information according to an exemplary embodiment of the present disclosure is schematically shown;

[0117] Figure 2 A schematic diagram of a structural example of a similarity calculation model according to an exemplary embodiment of the present disclosure is shown;

[0118] Figure 3 A schematic diagram of a structure example of a content generation model according to an exemplary embodiment of the present disclosure is shown;

[0119] Figure 4 A diagram schematically shows an example of a scene of lyrics information recording obtained according to an exemplary embodiment of the present disclosure;

[0120] Figure 5 A specific scenario example diagram schematically shows a method of determining original complete lyric information from candidate lyric information according to a first text similarity according to an exemplary embodiment of the present disclosure;

[0121] Figure 6 A diagram schematically shows an example of a scenario of matching an uncoded lyric line corresponding to a coded lyric line according to an example embodiment of the present disclosure;

[0122] Figure 7 A scene example diagram schematically shows a coded lyrics line before restoration and a lyrics line after restoration according to an exemplary embodiment of the present disclosure;

[0123] Figure 8 A block diagram schematically shows a device for processing lyrics information according to an exemplary embodiment of the present disclosure;

[0124] Fig. 9 Schematically illustrates a computer-readable storage medium for storing a processing method of lyrics information according to an exemplary embodiment of the present disclosure;

[0125] Fig.10 An electronic device for implementing a method for processing lyrics information according to an exemplary embodiment of the present disclosure is schematically shown.

[0126] In the drawings, the same or corresponding reference numerals represent the same or corresponding parts. DETAILED DESCRIPTION

[0127] The principles and spirit of the present disclosure will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are provided only to enable those skilled in the art to better understand and implement the present disclosure, and are not intended to limit the scope of the present disclosure in any way. On the contrary, these embodiments are provided to make the present disclosure more thorough and complete, and to fully convey the scope of the present disclosure to those skilled in the art.

[0128] Those skilled in the art will appreciate that the embodiments of the present disclosure may be implemented as a system, device, apparatus, method or computer program product. Therefore, the present disclosure may be specifically implemented in the following forms, namely: complete hardware, complete software (including firmware, resident software, microcode, etc.), or a combination of hardware and software.

[0129] According to an embodiment of the present disclosure, a method for processing lyrics information, a device for processing lyrics information, a computer-readable storage medium, and an electronic device are proposed.

[0130] In this document, any number of elements in the drawings is used for illustration rather than limitation, and any naming is used only for distinction and does not have any limiting meaning.

[0131] The principle and spirit of the present disclosure are explained in detail below with reference to several representative embodiments of the present disclosure. SUMMARY OF THE INVENTION

[0133] The applicant has found that with the increasing awareness of music copyright and strict content review, the phenomenon of lyrics coding is becoming more and more common. In the process of actual application, for the lyrics of some songs that are mistakenly coded, it is necessary to reply to the mistakenly coded lyrics to provide a better user experience. In some related technical schemes, when restoring the mistakenly coded lyrics, it can be achieved through the following technical schemes: one is to restore based on the data source provided by a professional music data provider; another is to restore the lyrics based on AIGC (Artificial Intelligence Generated Content); another is to manually restore the mistakenly coded lyrics by manual means; and another is to restore the mistakenly coded lyrics based on rule-based lyrics matching. Specifically:

[0134] For the first implementation method, the specific implementation idea is: first, obtain lyrics data from professional music data providers; among them, professional music data providers have a large amount of music resources and related data, and they can provide lyrics data to users through certain authorization methods; secondly, users can obtain original lyrics from these providers, and restore the lyrics that were mistakenly coded based on the original lyrics. However, this method has the following defects: on the one hand, in the actual application process, since some songs may not be able to obtain complete lyrics, other data sources need to be further supplemented; on the other hand, due to the complexity and diversity of music data and restrictions on copyright, professional music data providers may not be able to provide complete lyrics for all songs; for example, some niche songs or songs with complex copyright ownership may only be able to obtain partial lyrics information from the data provider, which affects the accuracy and completeness of lyrics coding recovery.

[0135] For the second implementation method, the specific implementation idea is: using the technology of artificial intelligence to generate content, by giving specific prompt words, let the model analyze and match the lyrics to find content similar to the coded lyrics, and try to restore the coded part. However, this method has the following defects: on the one hand, although the technology of artificial intelligence to generate content has certain innovation and application potential, it still has some limitations in the restoration of coded lyrics, such as the problem of hallucination; on the other hand, due to the complexity and uncertainty of the model, the results it generates may not be accurate and there is a possibility of error; on the other hand. The processing speed of artificial intelligence generated content is relatively slow, and it may not meet the real-time requirements for the restoration of a large number of song lyrics.

[0136] For the third implementation method, the specific implementation idea is: professionals restore the coded lyrics through manual modification; the professionals recorded here usually have rich music knowledge and lyrics understanding ability, and they can analyze and restore the coded lyrics one by one by combining their existing experience and knowledge with their understanding of the theme, style and singer of the song. However, this method has the following defects: extremely low efficiency and high cost; that is, in the actual application process, manual recovery requires a lot of manpower and time costs, and the efficiency is very low; therefore, for the needs of coded recovery of a large number of song lyrics, this method is almost not feasible and cannot meet the needs of actual application.

[0137] For the fourth implementation, the specific implementation idea is: formulate a series of rules to match lyrics; wherein, the rules recorded here can be set based on the number of words, format, theme and part of speech of the lyrics; in the actual application process, the coded lyrics can be matched with the rules to find lyrics that meet the rules as a reference for recovery. However, this method has the following defects: poor flexibility and low accuracy; that is, in the actual application process, the rule-based lyrics matching method is too rigid and cannot adapt to various complex lyrics situations; and, because the lyrics of different songs are very diverse and complex, it is difficult to accurately find matching lyrics by relying solely on fixed rules, which affects the effect of lyrics coding recovery.

[0138] Therefore, based on the current lyrics coding recovery technology, which has problems such as low song coverage, insufficient accuracy, slow speed, and insufficient support for long-tail songs, the example embodiment of the present disclosure provides a method for processing lyrics information, based on which the effect of lyrics coding recovery can be improved. At the same time, in the process of actual application, the specific implementation process of the method for processing lyrics information recorded in the example embodiment of the present disclosure can be divided into the following three parts: the first part is the collection of lyrics of coded songs, which collects song lyrics in a variety of ways to determine the source of the correctness of the lyrics coding; the second part is to restore the lyrics content, combining the lyrics matching algorithm and the ability of the large model to restore the coded lyrics; the third part is to determine whether the coded content is unprocessed, and finally determine whether the lyrics are coded by mistake through secondary risk control detection and word segmentation capabilities.

[0139] After introducing the basic principles of the present disclosure, various non-limiting embodiments of the present disclosure are described in detail below.

[0140] Exemplary Methods

[0141] The exemplary embodiment of the present disclosure first provides a method for processing lyrics information, which can be run on a terminal device, a server, a server cluster, a cloud server, etc. Of course, those skilled in the art can also run the method of the present disclosure on other platforms as required, and this exemplary embodiment does not specifically limit this. Figure 1 As shown, the method for processing lyrics information may include the following steps:

[0142] S110. Matching the original complete lyrics information corresponding to the coded lyrics information from a preset song database;

[0143] S120. According to the coded lyrics line in the coded lyrics information, matching the uncoded lyrics line corresponding to the coded lyrics line from the original complete lyrics information;

[0144] S130. Verify the accuracy of the coded content in the coded lyrics line according to the uncoded lyrics line;

[0145] S140. When it is determined that the coded content of the coded lyrics line has errors, the coded content of the coded lyrics line is restored according to the uncoded lyrics line.

[0146] In the above-mentioned method for processing lyric information, the original complete lyric information corresponding to the coded lyric information can be matched from a preset song database; then, based on the coded lyric lines in the coded lyric information, the uncoded lyric lines corresponding to the coded lyric lines are matched from the original complete lyric information; then, the accuracy of the coded content in the coded lyric lines is verified based on the uncoded lyric lines; finally, when it is determined that there is an error in the coded content of the coded lyric line, the coded content of the coded lyric line is restored based on the uncoded lyric line, without the need to manually restore the coded lyrics, thereby significantly reducing the problem of low recovery efficiency caused by manually restoring the coded lyrics, and reducing the problem of low accuracy of the recovery result caused by restoring the coded lyrics through artificial intelligence generated content, bringing a better experience to users.

[0147] Hereinafter, the method for processing lyrics information recorded in the exemplary embodiment of the present disclosure will be explained and illustrated in detail with reference to the accompanying drawings.

[0148] First, the terms involved in the exemplary embodiments of the present disclosure are explained and illustrated.

[0149] Lyrics Coding Recovery (LCR) refers to the process of repairing coded lyrics through technical means such as internal and external song matching. In this process, it is necessary to use a variety of technologies and algorithms to process and analyze lyrics data from different sources to obtain the original lyrics information and determine the correctness of the coding.

[0150] Lyrics Matching (LM): The process of selecting the lyrics closest to the current use case from lyrics from multiple sources; in actual application, lyrics matching involves calculating the similarity of lyrics texts. Commonly used methods include Jaccard similarity algorithm and edit distance algorithm. These algorithms can measure the similarity between different lyrics and find the most matching lyrics.

[0151] Mis-coding judgment: The process of judging whether the coded lyrics meet the correct coding requirements. In actual applications, the mis-coding judgment process needs to consider multiple factors, such as word segmentation, part-of-speech analysis, and sentiment analysis. It also needs to compare the coding conditions of different data sources and sensitive word libraries to comprehensively judge whether the lyrics are mis-coded.

[0152] Lyrics Similarity Calculation (LSC): A calculation method used to measure the similarity between different lyrics. In addition to the Jaccard similarity algorithm and the edit distance algorithm, other text similarity calculation methods may also be involved. These methods are based on different principles and formulas, and analyze and compare the words or characters in the lyrics text to obtain the similarity value.

[0153] Secondary risk control detection: During the process of recovering lyrics after coding, the obtained lyrics are subjected to risk control detection again. By calling the risk control interface to perform secondary sensitive word detection and comparing sensitive word libraries, the lyrics content is ensured to meet security requirements and avoid violations or sensitive information.

[0154] Word segmentation tool: a tool used to segment lyrics text. Specifically, in the process of judging the mis-coding, the lyrics can be segmented using a large model and a three-party word segmentation tool, so as to better analyze the lyrics content and judge the correctness of the coding.

[0155] Sensitive word library: a database used to store sensitive words. During the process of recovering lyrics after coding, it is necessary to compare the sensitive word library to determine whether the lyrics contain sensitive information; of course, the sensitive word library is also used in the secondary risk control detection process to ensure that the lyrics content meets security requirements.

[0156] Jaccard similarity algorithm: Jaccard similarity is a method used to measure the similarity between two sets, and is often used in text similarity calculations. The calculation principle of the Jaccard similarity algorithm can be implemented based on set theory. Specifically, in practical applications, the calculation formula for Jaccard similarity is: J(A,B) = |A∩B| / |A∪B|, where A and B are two sets, |A∩B| represents the size of the intersection of the two sets, and |A∪B| represents the size of the union of the two sets. Through this calculation formula, Jaccard similarity can be interpreted as the ratio of the size of the intersection of the two sets to the size of the union. The value range of Jaccard similarity is between 0 and 1, and the value is 1 when the two sets are exactly the same, and 0 when they are completely different. At the same time, in the calculation of text similarity, the words or n-grams in the text can be regarded as elements in the set, and then the Jaccard similarity is used to measure the similarity between the two texts.

[0157] Edit distance algorithm: used to calculate the edit distance between two texts, that is, how many editing operations (insertion, deletion, replacement) are required to convert one text into another.

[0158] AIGC: Artificial Intelligence Generated Content, artificial intelligence generates content; specifically, in the actual application process, the Prompt prompt can be used to let the large model ignore the differences in text format and the star part, match the similarity between the input sample and the candidate lyrics, and increase the matching weight for the front and back lines of the * part of the lyrics.

[0159] Next, the similarity calculation model involved in the exemplary embodiment of the present disclosure is explained and illustrated. Figure 2 As shown, the similarity calculation model described herein may include a first input layer 210, a first embedding mapping layer 220, a similarity comparison layer 230, a row weight matching layer 240, a similarity calculation layer 250 and a first output layer 260; wherein, the functions required to be performed by the different functional layers described herein in the similarity calculation process will be described in detail later and will not be further elaborated here.

[0160] Further, the content generation model involved in the exemplary embodiment of the present disclosure is explained and illustrated. Figure 3 As shown, the content generation model described herein may include a second input layer 310, a second embedding mapping layer 320, a coding layer 330, a hybrid expert model layer 340, and a second output layer. The functions required to be performed by the different functional layers described herein during the similarity calculation process will be described in detail later, and will not be further described here.

[0161] The following will explain and illustrate the specific collection process of the original complete lyrics information recorded in the exemplary embodiment of the present disclosure. Specifically, in the actual application process, in order to match the original complete lyrics information corresponding to the coded lyrics information from the song database, it is necessary to first collect the original complete lyrics information; further, in the process of collecting the original complete lyrics information, it is necessary to first submit a lyrics collection task; at the same time, after the lyrics collection task is submitted, the original complete lyrics information will be obtained based on different channels; at the same time, it should be supplemented here that the premise of obtaining the original complete lyrics information is to obtain the core information of the song, such as the song name, artist name, album name, etc.; therefore, when it is necessary to collect the original complete lyrics information of a song, a lyrics collection request can be generated based on the song name, artist name, album name and other information, and then the lyrics collection request is sent to different servers for the original complete lyrics information collection; for example, the original complete lyrics information disclosed can be collected based on the robots of the website; for another example, the source lyrics data provided by the song manufacturer can also be obtained; for another example, the original complete lyrics information can also be obtained from a third-party paid lyrics database. Furthermore, after obtaining the original complete lyrics information from different sources, the original complete lyrics information from different sources of the same song can be written into the same record for subsequent matching processes; among which, the scene example diagram of the recording of the obtained lyrics information can be referred to Figure 4 shown.

[0162] The following will be combined Figure 2-Figure 4 right Figure 1 The processing method of the lyrics information shown in the following is further explained and illustrated. Specifically:

[0163] In S110, the original complete lyrics information corresponding to the coded lyrics information is matched from a preset song database.

[0164] Specifically, the specific matching process of the original complete lyrics information can be implemented in the following manner: first, obtain the coded lyrics information of the current song and the current song attribute information of the current song; wherein the current song attribute information includes at least one of the current song name, the current song singer, and the song album name to which the current song belongs; secondly, match the candidate lyrics information corresponding to the current song from the preset song database according to the current song attribute information, obtain a first matching result, and determine the original complete lyrics information corresponding to the coded lyrics information according to the first matching result. That is to say, in the actual application process, the matching of the original complete lyrics information can be directly based on the song name, the current song singer, and the song album name; for example, the song singer can be determined first, then the song album name can be determined, and finally the song name can be determined; if all can be matched, the first matching result is determined to be non-empty; if any one item cannot be matched, the first matching result is empty.

[0165] In an exemplary embodiment, determining the original complete lyric information corresponding to the coded lyric information based on the first matching result can be achieved in the following manner: when it is determined that the first matching result is not empty, calculating the first text similarity between the coded lyric information and the candidate lyric information, and determining the original complete lyric information from the candidate lyric information based on the first text similarity; when it is determined that the first matching result is empty, matching the associated songs and lyric context corresponding to the current song from the preset song database according to the current song attribute information to obtain a second matching result; if the second matching result is not empty, determining the original complete lyric information based on the associated songs and lyric context, and if the second matching result is empty, generating the original complete lyric information based on a preset content generation model. That is to say, in the actual application process, if the corresponding original complete lyrics information can be matched from the song database, then the one with the highest similarity can be directly selected from the matched original complete lyrics information as the original complete lyrics information of the song; if the original complete lyrics information cannot be matched, the corresponding related songs and lyrics context can be determined based on the song singer, song name and album to which the song belongs, and then the corresponding original complete lyrics information can be determined based on the related songs and lyrics context; further, if the corresponding related songs and lyrics context cannot be matched, the corresponding information can be directly input into the content generation model to generate the original complete lyrics information in an AIGC-based manner.

[0166] In an exemplary embodiment, calculating the first text similarity between the coded lyric information and the candidate lyric information can be achieved in the following ways: a first implementation method is: calculating the first text similarity between the coded lyric information and the candidate lyric information based on a preset similarity algorithm; a second implementation method is: calculating the first text similarity between the coded lyric information and the candidate lyric information based on a preset edit distance algorithm; a third implementation method is: calculating the first text similarity between the coded lyric information and the candidate lyric information based on a preset similarity calculation model.

[0167] The following will further explain and illustrate the calculation methods of different first text similarities. Specifically:

[0168] For the first implementation, the specific calculation process of the first text similarity can be implemented as follows: first, the coded lyrics information is segmented to obtain the first segmentation result, and the first segmentation set is constructed based on the first segmentation result; second, the candidate lyrics information is segmented to obtain the second segmentation result, and the second segmentation set is constructed based on the second segmentation result; then, the intersection word set and the union word set between the first segmentation set and the second segmentation set are calculated, and the first text similarity between the coded lyrics information and the candidate lyrics information is determined according to the intersection word set and the union word set. That is, in the actual application process, the first segmentation set A and the second segmentation combination B can be constructed first; then, the intersection A∩B between the first segmentation set A and the second segmentation combination B, and the union A∪B between the first segmentation set A and the second segmentation combination B are calculated; finally, |A∩B| / |A∪B| are calculated to obtain the first text similarity.

[0169] For the second implementation, the specific calculation process of the first text similarity can be implemented as follows: First, the preset edit distance algorithm recorded above can be a single character edit operation, which can include an insert operation, a delete operation, and a replace operation, etc. For example, the two character sequences "XAAB7818" and "XAAB7098" require the minimum single character edit operations to convert "XAAB7818" to "XAAB7098": 8 becomes 0, 1 becomes 9, so the edit distance between the two license plates "XAAB7818" and "XAAB7098" is 2.

[0170] Generally speaking, the smaller the edit distance between two strings, the more similar they are; if two strings are equal, their edit distance is 0; at the same time, in the actual application process, it is assumed that the LevenshteinDistant (character edit distance) of two strings a and b is expressed as:

[0171] lev a,b =(|a|,|b|)

[0172] Among them, |a| and |b| correspond to the length of the string and string b respectively, so lev a,b =(|a|,|b|), which can be described in the following data language:

[0173]

[0174] Among them, when a i =b j When , the distance between the i-th character in string a and the j-th character in string b is zero; when a i ≠b j When , the distance between the i-th character in string a and the j-th character in string b is 1; lev a,b Represents the distance from the first i characters of a to the first j characters of b, where relative to the change from a to b string:

[0175] lev a,b (i-1,j)+1, means a deletes one character to match b;

[0176] lev a,b (i,j-1)+1, means a adds a character to match b;

[0177] lev a,b (i-1,j-1)+1, indicating a match or a mismatch, depending on whether the symbols are the same.

[0178] Under the premise of the above-mentioned contents, single-character editing operation can be performed on the coded lyric information and the candidate lyric information to obtain the character editing distance between the coded lyric information and the candidate lyric information, and then the first text similarity can be obtained based on the character editing distance of each candidate lyric information.

[0179] For the third implementation method, the specific calculation process of the first text similarity can be implemented as follows: first, based on the first embedding mapping layer, calculate the first semantic text vector of each line of lyrics in the coded lyrics information and the second semantic text vector of each line of lyrics in the candidate lyrics information; wherein, the first embedding mapping layer recorded here may include an Embedding embedding mapping layer and / or a Bert embedding mapping layer; based on the similarity comparison layer, calculate the similarity between the first semantic text vector and the second semantic text vector to obtain the line text similarity; based on the line weight matching layer, match the line weight value for the first semantic text vector according to the first line position of the lyrics corresponding to the first semantic text vector in the coded lyrics information; based on the similarity calculation layer, weighted sum the line text similarity and the line weight value corresponding to the line text similarity to obtain the first text similarity. That is, in the actual application process, based on the Prompt word, the large model can be made to ignore the differences in text format and the star-marked part (that is, the coded part), and then match the input sample (that is, the coded lyrics information) and the candidate lyrics information to obtain the first text similarity between the coded lyrics information and the candidate lyrics information; at the same time, in order to further improve the accuracy of the first text similarity obtained, the coded part of the lyrics can also be combined to increase the matching weight of the previous and following lines adjacent to the coded part of the lyrics, thereby achieving the purpose of weakening the proportion of the coded part of the lyrics in the overall similarity calculation process. Among them, the specific scenario example diagram of determining the original complete lyrics information from the candidate lyrics information based on the first text similarity can be referred to. Figure 5 shown.

[0180] In an exemplary embodiment, the original complete lyric information is generated according to the preset content generation model, which can be achieved in the following manner: text preprocessing is performed on the coded lyric information to obtain standard lyric information, and lyrics context information is generated according to the preset content parameter prompt information corresponding to the coded lyric information and the current song attribute information; word embedding processing is performed on the standard lyric information based on the second embedding mapping layer to obtain lyric text features, and embedding mapping processing is performed on the lyric context information to obtain a context marker sequence, and encoding processing is performed on the lyric text features and the context marker sequence based on the encoding layer to obtain a context overall representation; based on the hybrid expert model layer, the context marker sequence and the context overall representation are predicted to obtain a text prediction result, and the text prediction result is matched in the first preset sensitive word library to obtain a third matching result; if the third matching result is that the text prediction result exists in the first preset sensitive word library, the original complete lyric information is generated according to the text prediction result. Specifically, the text preprocessing recorded here may include but is not limited to removing line break logic, removing special symbols (such as spaces) and time prefixes, etc.; then, lyrics prediction is performed based on the content generation model; further, in the process of predicting the context marker sequence and the overall context representation based on the hybrid expert model layer to obtain the text prediction result, it can be achieved from the two dimensions of content fit and the overall style fit, or other dimensions can be set, and this example does not impose any special restrictions on this; further, in order to better improve the accuracy of the original complete lyrics information obtained, it is also necessary to determine whether the obtained text prediction results exist in the sensitive word library; that is, if the predicted coded lyrics exist in the sensitive word library, it is determined that the prediction result is accurate; otherwise, it is necessary to manually restore the coded lyrics.

[0181] In S120, according to the coded lyrics line in the coded lyrics information, the uncoded lyrics line corresponding to the coded lyrics line is matched from the original complete lyrics information.

[0182] Specifically, the specific process of determining the uncoded lyrics line is: determine the first line position of the coded lyrics line in the coded lyrics information, and match the uncoded lyrics line corresponding to the coded lyrics line from the original complete lyrics information according to the first line position. For example, if the coded lyrics line is in the third line of the coded lyrics information, the uncoded lyrics line can be extracted from the third line of the original complete lyrics information; the specific matching process can be referred to. Figure 6 shown.

[0183] In S130, the accuracy of the coded content in the coded lyrics line is verified based on the uncoded lyrics line.

[0184] Specifically, the accuracy verification process of the coded content in the coded lyrics line can be achieved in the following manner: compare the uncoded lyrics line with the coded lyrics line to obtain the coded content in the coded lyrics line, and input the uncoded lyrics line into a preset coded content prediction model to obtain a coding prediction result; verify the accuracy of the coding prediction result, and when it is determined that the coding prediction result is accurate, compare the coding prediction result with the coded content in the same lyrics line as the coding prediction result to verify the accuracy of the coded content; wherein, if the coding prediction result and the coded content in the same lyrics line as the coding prediction result are consistent, then it is determined that the accuracy of the coded content of the coded lyrics line is correct; if the coding prediction result and the coded content in the same lyrics line as the coding prediction result are inconsistent, then it is determined that the accuracy of the coded content of the coded lyrics line is incorrect. Wherein, the coded content prediction model recorded here can be a deep neural network model (such as a convolutional neural network model or a recurrent neural network model, etc.).

[0185] In an example embodiment, the specific implementation process of verifying the accuracy of the coding prediction result is: verifying the accuracy of the coding prediction result according to the second line position of each coding prediction result in the uncoded lyrics and the remaining lyrics in the uncoded lyrics except the coding prediction result. That is, in the actual application process, first, it can be determined whether the number of lyrics lines before and after restoration is consistent with the original complete lyrics information; if consistent, it is determined whether the uncoded content has changed; if not, it is determined whether the coded content is in the sensitive word library, etc.; if any one of the items is no, it is determined that the coding prediction result is inaccurate, and the coded lyrics need to be restored manually; it is also necessary to further explain here that in order to improve the accuracy of the judgment result of whether the coding prediction result is accurate, the uncoded lyrics line can be continuously input into the preset coding content prediction model for multiple times, so as to obtain multiple coding prediction results, and then the accuracy of the coding prediction result is judged based on the comprehensive judgment of each coding prediction result.

[0186] In S140, when it is determined that the coded content of the coded lyrics line has errors, the coded content of the coded lyrics line is restored according to the uncoded lyrics line.

[0187] Specifically, the specific recovery process of the coded content can be achieved in the following way: identify the coded content of the coded lyrics line with errors to obtain a content identification result, and obtain the lyrics to be replaced from the uncoded lyrics line based on the third line position of the coded lyrics line to which the content identification result belongs in the coded lyrics information and the identification position of the content identification result in the coded lyrics line; replace the content identification result based on the lyrics to be replaced to achieve recovery of the coded content of the coded lyrics line. That is, in the actual application process, if there is an error in the coded content, it can be directly restored based on the uncoded lyrics line; wherein, the coded lyrics line before recovery and the lyrics line after recovery can be specifically referred to. Figure 7 shown.

[0188] In an example embodiment, after the coded lyrics are restored, it is necessary to perform secondary risk control on the restored lyrics; specifically, this can be achieved in the following way: first, the coded content and the uncoded lyrics lines corresponding to the coded content are input into a preset risk control detection model to obtain a sensitive content detection result, and the coded content is subjected to secondary risk control detection based on the sensitive content detection result; secondly, if the secondary risk control detection result is that the coded content is safe content, a content recovery request is generated based on the restored coded content, and the content recovery request is pushed to the client where the operation and maintenance personnel are located, so as to perform online recovery of the coded content. Furthermore, if the secondary risk control detection result is that the coded content is sensitive content, the uncoded lyrics line is segmented to obtain a second segmentation result, and the second segmentation result and the uncoded lyrics line are input into a preset sensitive word prediction model to obtain a first sensitive content detection result; the segmentation result and the uncoded lyrics are input into a second preset sensitive word library to obtain a second sensitive content detection result, and the segmentation result and the uncoded lyrics are input into a third preset sensitive word library to obtain a third sensitive content detection result; according to the first sensitive content detection result, the second sensitive content detection result and the third sensitive content detection result, the target sensitive word is determined from the segmentation result, and it is determined whether the target sensitive word is consistent with the coded content; if the target sensitive word is consistent with the coded content, the coded content in the coded lyrics line is kept in a coded state; otherwise, a content recovery request is generated according to the restored coded content, and the content recovery request is pushed to the client where the operation and maintenance personnel are located to perform online recovery of the coded content. Specifically, in the actual application process, since the risk control detection technology has been iterating, the accuracy rate has gradually improved; therefore, in the process of judging the error coding, a round of risk control detection will be performed first. If the secondary risk control output result is not coded, it means that the original coding is a misjudgment, and the lyrics content can be directly restored. Of course, if the secondary risk control output result is also coded, it will be segmented with the help of large model segmentation and third-party segmentation tools. If the content after segmentation and the coded content are not the same word, it means that the lyrics are mistakenly coded; and, considering that the accuracy of the large model and different third-party segmentation tools is related to their training methods and training corpus, it is not necessarily correct to look at one of them alone. Therefore, the selection will use 3 tools for segmentation at the same time, and the results will be selected from 3 to 2, which can ensure the accuracy of the segmentation results to a large extent.

[0189] At this point, the method for processing lyric information recorded in the example embodiment of the present disclosure has been fully realized. Based on the above-mentioned contents, it can be known that the method for processing lyric information recorded in the example embodiment of the present disclosure has at least the following advantages: on the one hand, the song coverage rate is improved; that is, the example embodiment of the present disclosure can comprehensively adopt a variety of data sources and technical means, such as comparison of existing original lyrics, data sources of professional music data providers, and large models, which greatly improves the range of songs that can be processed; at the same time, the coding and recovery of lyrics for songs with different copyright situations, storage status, and song popularity can be effectively covered; moreover, whether it is a common song or a long-tail song, there is a corresponding processing method, which is no longer limited to a few songs with original lyrics stored, thereby greatly improving the overall song coverage rate and being able to meet the needs of more users for lyrics recovery; on the other hand, the accuracy is improved; that is, in terms of miscoding judgment, the example embodiment of the present disclosure adopts a solution of secondary risk control detection + re-segmentation and matching of sensitive word libraries to analyze and verify the lyrics from multiple dimensions; and, by combining different segmentation tools for segmentation, and The results are voted on to determine the final result, reducing the errors that may be caused by a single method; at the same time, the secondary risk control detection can further check for possible sensitive information risks, ensuring that the lyrics content not only meets the requirements in terms of coding accuracy, but also can be guaranteed in terms of content security; further, based on the comprehensive judgment method, the accuracy of judging whether the lyrics are correctly coded is improved, avoiding the occurrence of misjudgment, and providing users with more accurate lyrics recovery results; on the other hand, the speed is moderate; that is, the example embodiment of the present disclosure comprehensively considers the characteristics of various methods, and the overall solution can meet actual needs in terms of speed; and this comprehensive solution design ensures a certain processing efficiency when processing a large number of song lyrics coding recovery tasks, and will not sacrifice accuracy in pursuit of speed, and can provide users with lyrics recovery services within a reasonable time; further, it reduces labor costs; that is, compared with the traditional manual method of restoring lyrics, the comprehensive solution of the example embodiment of the present disclosure greatly reduces labor costs. Manual recovery is not only extremely inefficient, but also requires a lot of time and energy from professionals. Moreover, the example embodiments disclosed herein can complete the task of coded lyrics recovery with less human intervention through automated processes and algorithms, freeing up human resources from tedious manual recovery work and allowing them to be used for other more valuable work links, thereby improving the operational efficiency and economic benefits of the entire platform.

[0190] Exemplary Devices

[0191] After introducing the method for processing lyrics information in the exemplary embodiment of the present disclosure, next, refer to Figure 8 A lyric information processing device according to an exemplary embodiment of the present disclosure is explained and described.

[0192] Specifically, refer to Figure 8 As shown, the lyric information processing device may include a complete lyric information matching module 810, an uncoded lyric line matching module 820, an accuracy checking module 830, and a coded content recovery module 840. Among them:

[0193] The complete lyrics information matching module 810 may be used to match the original complete lyrics information corresponding to the coded lyrics information from a preset song database;

[0194] The uncoded lyrics line matching module 820 may be used to match the uncoded lyrics line corresponding to the coded lyrics line from the original complete lyrics information according to the coded lyrics line in the coded lyrics information;

[0195] The accuracy verification module 830 may be used to verify the accuracy of the coded content in the coded lyrics line according to the uncoded lyrics line;

[0196] The coded content recovery module 840 may be used to recover the coded content of the coded lyrics line according to the uncoded lyrics line when it is determined that the coded content of the coded lyrics line has errors.

[0197] In an exemplary embodiment of the present disclosure, matching the original complete lyric information corresponding to the coded lyric information from a preset song database includes: obtaining the coded lyric information of the current song and the current song attribute information of the current song; wherein the current song attribute information includes at least one of the current song name, the current song singer, and the song album name to which the current song belongs; matching the candidate lyric information corresponding to the current song from the preset song database according to the current song attribute information to obtain a first matching result, and determining the original complete lyric information corresponding to the coded lyric information according to the first matching result.

[0198] In an exemplary embodiment of the present disclosure, the original complete lyric information corresponding to the coded lyric information is determined according to the first matching result, including: when it is determined that the first matching result is not empty, calculating the first text similarity between the coded lyric information and the candidate lyric information, and determining the original complete lyric information from the candidate lyric information according to the first text similarity; when it is determined that the first matching result is empty, matching the associated songs and lyric context corresponding to the current song from the preset song database according to the current song attribute information to obtain a second matching result; if the second matching result is not empty, determining the original complete lyric information according to the associated songs and lyric context, and if the second matching result is empty, generating the original complete lyric information according to a preset content generation model.

[0199] In an exemplary embodiment of the present disclosure, calculating the first text similarity between the coded lyric information and the candidate lyric information includes: calculating the first text similarity between the coded lyric information and the candidate lyric information based on a preset similarity algorithm; and / or calculating the first text similarity between the coded lyric information and the candidate lyric information based on a preset edit distance algorithm; and / or calculating the first text similarity between the coded lyric information and the candidate lyric information based on a preset similarity calculation model.

[0200] In an exemplary embodiment of the present disclosure, a first text similarity between the coded lyric information and the candidate lyric information is calculated based on a preset similarity algorithm, including: performing word segmentation processing on the coded lyric information to obtain a first word segmentation result, and constructing a first word segmentation set based on the first word segmentation result; performing word segmentation processing on the candidate lyric information to obtain a second word segmentation result, and constructing a second word segmentation set based on the second word segmentation result; calculating an intersection word set and a union word set between the first word segmentation set and the second word segmentation set, and determining the first text similarity between the coded lyric information and the candidate lyric information based on the intersection word set and the union word set.

[0201] In an exemplary embodiment of the present disclosure, the preset similarity calculation model includes a first embedding mapping layer, a similarity comparison layer, a row weight matching layer and a similarity calculation layer; wherein, the first text similarity between the coded lyrics information and the candidate lyrics information is calculated based on the preset similarity calculation model, including: calculating the first semantic text vector of each line of lyrics in the coded lyrics information and the second semantic text vector of each line of lyrics in the candidate lyrics information based on the first embedding mapping layer; calculating the similarity between the first semantic text vector and the second semantic text vector based on the similarity comparison layer to obtain the row text similarity; matching the row weight value for the first semantic text vector according to the first row position of the lyrics corresponding to the first semantic text vector in the coded lyrics information based on the row weight matching layer; performing weighted summation of the row text similarity and the row weight value corresponding to the row text similarity based on the similarity calculation layer to obtain the first text similarity.

[0202] In an exemplary embodiment of the present disclosure, the preset content generation model includes a second embedding mapping layer, an encoding layer and a hybrid expert model layer; wherein, generating the original complete lyrics information according to the preset content generation model includes: performing text preprocessing on the coded lyrics information to obtain standard lyrics information, and generating lyrics context information according to the preset content parameter prompt information corresponding to the coded lyrics information and the current song attribute information; performing word embedding processing on the standard lyrics information based on the second embedding mapping layer to obtain lyrics text features, and performing embedding mapping processing on the lyrics context information to obtain a context marker sequence, and encoding processing on the lyrics text features and the context marker sequence based on the encoding layer to obtain a context overall representation; predicting the context marker sequence and the context overall representation based on the hybrid expert model layer to obtain a text prediction result, and matching the text prediction result in a first preset sensitive word library to obtain a third matching result; if the third matching result is that the text prediction result exists in the first preset sensitive word library, then generating the original complete lyrics information according to the text prediction result.

[0203] In an exemplary embodiment of the present disclosure, according to the coded lyrics line in the coded lyrics information, matching the uncoded lyrics line corresponding to the coded lyrics line from the original complete lyrics information includes: determining the first line position of the coded lyrics line in the coded lyrics information, and according to the first line position, matching the uncoded lyrics line corresponding to the coded lyrics line from the original complete lyrics information.

[0204] In an exemplary embodiment of the present disclosure, the accuracy of the coded content in the coded lyrics line is verified based on the uncoded lyrics line, including: comparing the uncoded lyrics line with the coded lyrics line to obtain the coded content in the coded lyrics line, and inputting the uncoded lyrics line into a preset coded content prediction model to obtain a coding prediction result; verifying the accuracy of the coding prediction result, and when it is determined that the coding prediction result is accurate, comparing the coding prediction result with the coded content in the same lyrics line as the coding prediction result to verify the accuracy of the coded content; wherein, if the coding prediction result and the coded content in the same lyrics line as the coding prediction result are consistent, then it is determined that the accuracy of the coded content of the coded lyrics line is correct; if the coding prediction result and the coded content in the same lyrics line as the coding prediction result are inconsistent, then it is determined that there is an error in the accuracy of the coded content of the coded lyrics line.

[0205] In an exemplary embodiment of the present disclosure, the accuracy of the coding prediction results is verified, including: verifying the accuracy of the coding prediction results based on the second line position of each coding prediction result in the uncoded lyrics and the remaining lyrics in the uncoded lyrics except the coding prediction results.

[0206] In an exemplary embodiment of the present disclosure, the coded content of the coded lyrics line is restored based on the uncoded lyrics line, including: identifying the coded content of the coded lyrics line with errors to obtain a content identification result, and obtaining lyrics to be replaced from the uncoded lyrics line based on the third line position of the coded lyrics line to which the content identification result belongs in the coded lyrics information and the identification position of the content identification result in the coded lyrics line; replacing the content identification result based on the lyrics to be replaced to achieve recovery of the coded content of the coded lyrics line.

[0207] In an exemplary embodiment of the present disclosure, the lyric information processing device further includes:

[0208] A secondary risk control detection module is used to input the coded content and the uncoded lyrics corresponding to the coded content into a preset risk control detection model to obtain a sensitive content detection result, and perform a secondary risk control detection on the coded content according to the sensitive content detection result;

[0209] The first content recovery request generating module is used to generate a content recovery request according to the recovered coded content if the secondary risk control detection result is that the coded content is safe content, and push the content recovery request to the client where the operation and maintenance personnel are located, so as to restore the coded content online.

[0210] In an exemplary embodiment of the present disclosure, the lyric information processing device further includes:

[0211] A first sensitive content detection module, configured to, if the secondary risk control detection result is that the coded content is sensitive content, perform word segmentation processing on the uncoded lyrics line to obtain a second word segmentation result, and input the second word segmentation result and the uncoded lyrics line into a preset sensitive word prediction model to obtain a first sensitive content detection result;

[0212] A second sensitive content detection module, configured to input the word segmentation result and the uncoded lyrics into a second preset sensitive word library to obtain a second sensitive content detection result, and input the word segmentation result and the uncoded lyrics into a third preset sensitive word library to obtain a third sensitive content detection result;

[0213] a target sensitive word determination module, configured to determine a target sensitive word from the word segmentation result according to the first sensitive content detection result, the second sensitive content detection result, and the third sensitive content detection result, and to determine whether the target sensitive word is consistent with the coded content;

[0214] The second content recovery request generation module is used to keep the coded content in the coded lyrics line in a coded state if the target sensitive word is consistent with the coded content; otherwise, generate a content recovery request based on the restored coded content, and push the content recovery request to the client where the operation and maintenance personnel are located to restore the coded content online.

[0215] Exemplary Storage Media

[0216] After introducing the method for processing lyrics information and the device for processing lyrics information according to the exemplary embodiment of the present disclosure, Fig. 9 A storage medium according to an exemplary embodiment of the present disclosure is described.

[0217] refer to Fig. 9 As shown, a program product 900 for implementing the above method according to an embodiment of the present disclosure is described, which can adopt a portable compact disk read-only memory (CD-ROM) and include program code, and can be run on a terminal device, such as a personal computer. However, the program product of the present disclosure is not limited thereto.

[0218] The program product may use any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0219] Computer readable signal media may include data signals propagated in baseband or as part of a carrier wave, wherein readable program code is carried. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Readable signal media may also be any readable medium other than a readable storage medium.

[0220] Program code for performing the operations of the present disclosure may be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java, C++, etc., and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, partially on a remote computing device, or entirely on a remote computing device or server. In the case of a remote computing device, the remote computing device may be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN).

[0221] Exemplary Electronic Devices

[0222] After introducing the storage medium of the exemplary embodiment of the present disclosure, next, reference is made to Fig.10 An electronic device according to an exemplary embodiment of the present disclosure is described.

[0223] Fig.10 The electronic device 1000 shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0224] like Fig.10 As shown, the electronic device 1000 is in the form of a general computing device. The components of the electronic device 1000 may include but are not limited to: the at least one processing unit 1010, the at least one storage unit 1020, a bus 1030 connecting different system components (including the storage unit 1020 and the processing unit 1010), and a display unit 1040.

[0225] The storage unit 1020 stores program codes, which can be executed by the processing unit 1010, so that the processing unit 1010 performs the steps according to various exemplary embodiments of the present disclosure described in the above “Exemplary Method” section of this specification. For example, the processing unit 1010 can perform the following steps: Figure 1 Steps S110-S140 shown in FIG.

[0226] The storage unit 1020 may include a volatile storage unit, such as a random access storage unit (RAM) 10201 and / or a cache storage unit 10202 , and may further include a read-only storage unit (ROM) 10203 .

[0227] The storage unit 1020 may also include a program / utility 10204 having a set (at least one) of program modules 10205, such program modules 10205 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment.

[0228] The bus 1030 may include a data bus, an address bus, and a control bus.

[0229] The electronic device 1000 may also communicate with one or more external devices 1100 (e.g., keyboards, pointing devices, Bluetooth devices, etc.) through an input / output (I / O) interface 1050. Furthermore, the electronic device 1000 may also communicate with one or more networks (e.g., local area networks (LANs), wide area networks (WANs), and / or public networks, such as the Internet) through a network adapter 1060. As shown, the network adapter 1060 communicates with other modules of the electronic device 1000 through a bus 1030. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in conjunction with the electronic device 1000, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, etc.

[0230] It should be noted that although several modules or submodules of the pop-up window processing device are mentioned in the above detailed description, this division is merely exemplary and not mandatory. In fact, according to the embodiments of the present disclosure, the features and functions of two or more units / modules described above can be concretized in one unit / module. Conversely, the features and functions of one unit / module described above can be further divided into multiple units / modules for concretization.

[0231] In addition, although the operations of the disclosed method are described in a specific order in the drawings, this does not require or imply that the operations must be performed in this specific order, or that all the operations shown must be performed to achieve the desired results. Additionally or alternatively, some steps may be omitted, multiple steps may be combined into one step, and / or one step may be decomposed into multiple steps.

[0232] Although the spirit and principle of the present disclosure have been described with reference to several specific embodiments, it should be understood that the present disclosure is not limited to the disclosed specific embodiments, and the division of various aspects does not mean that the features in these aspects cannot be combined to benefit, and such division is only for the convenience of expression. The present disclosure is intended to cover various modifications and equivalent arrangements included in the spirit and scope of the attached claims.

Claims

1. A method for processing lyrics information, characterized in that: include: Matching the original complete lyrics information corresponding to the coded lyrics information from a preset song database; According to the coded lyrics line in the coded lyrics information, matching the uncoded lyrics line corresponding to the coded lyrics line from the original complete lyrics information; Verifying the accuracy of the coded content in the coded lyrics line according to the uncoded lyrics line; When it is determined that the coded content of the coded lyrics line has errors, the coded content of the coded lyrics line is restored according to the uncoded lyrics line.

2. The method for processing lyrics information according to claim 1, characterized in that: Matching the original complete lyrics information corresponding to the coded lyrics information from the preset song database, including: Obtaining coded lyrics information of a current song and current song attribute information of the current song; wherein the current song attribute information includes at least one of the current song title, the current song singer, and the song album name to which the current song belongs; According to the current song attribute information, candidate lyrics information corresponding to the current song is matched from a preset song database to obtain a first matching result, and the original complete lyrics information corresponding to the coded lyrics information is determined according to the first matching result.

3. The method for processing lyrics information according to claim 2, characterized in that: Determining original complete lyrics information corresponding to the coded lyrics information according to the first matching result includes: When it is determined that the first matching result is not empty, calculating a first text similarity between the coded lyrics information and the candidate lyrics information, and determining the original complete lyrics information from the candidate lyrics information according to the first text similarity; When it is determined that the first matching result is empty, matching the associated songs and lyrics context corresponding to the current song from the preset song database according to the current song attribute information to obtain a second matching result; If the second matching result is not empty, the original complete lyrics information is determined according to the associated song and the lyrics context; if the second matching result is empty, the original complete lyrics information is generated according to a preset content generation model.

4. The method for processing lyrics information according to claim 3, characterized in that: Calculating a first text similarity between the coded lyrics information and the candidate lyrics information includes: Calculating a first text similarity between the coded lyrics information and the candidate lyrics information based on a preset similarity algorithm; and / or Calculating a first text similarity between the coded lyrics information and the candidate lyrics information based on a preset edit distance algorithm; and / or The first text similarity between the coded lyrics information and the candidate lyrics information is calculated based on a preset similarity calculation model.

5. The method for processing lyrics information according to claim 4, characterized in that: Calculating a first text similarity between the coded lyrics information and the candidate lyrics information based on a preset similarity algorithm includes: Performing word segmentation processing on the coded lyrics information to obtain a first word segmentation result, and constructing a first word segmentation set based on the first word segmentation result; Performing word segmentation processing on the candidate lyrics information to obtain a second word segmentation result, and constructing a second word segmentation set based on the second word segmentation result; An intersection word set and a union word set between the first word segmentation set and the second word segmentation set are calculated, and a first text similarity between the coded lyric information and the candidate lyric information is determined based on the intersection word set and the union word set.

6. The method for processing lyrics information according to claim 4, characterized in that: The preset similarity calculation model includes a first embedding mapping layer, a similarity comparison layer, a row weight matching layer and a similarity calculation layer; The step of calculating the first text similarity between the coded lyrics information and the candidate lyrics information based on a preset similarity calculation model includes: Calculate, based on the first embedding mapping layer, a first semantic text vector for each line of lyrics in the coded lyrics information and a second semantic text vector for each line of lyrics in the candidate lyrics information; Calculate the similarity between the first semantic text vector and the second semantic text vector based on the similarity comparison layer to obtain line text similarity; Matching a row weight value for the first semantic text vector based on the row weight matching layer according to the first row position of the lyrics corresponding to the first semantic text vector in the coded lyrics information; Based on the similarity calculation layer, a weighted sum is performed on the line text similarity and the line weight value corresponding to the line text similarity to obtain the first text similarity.

7. The method for processing lyrics information according to claim 3, characterized in that: The preset content generation model includes a second embedding mapping layer, a coding layer and a hybrid expert model layer; The method of generating the original complete lyrics information according to the preset content generation model includes: Performing text preprocessing on the coded lyrics information to obtain standard lyrics information, and generating lyrics context information according to preset content parameter prompt information corresponding to the coded lyrics information and current song attribute information; Based on the second embedding mapping layer, the standard lyrics information is subjected to word embedding processing to obtain lyrics text features, and the lyrics context information is subjected to embedding mapping processing to obtain a context marker sequence, and based on the encoding layer, the lyrics text features and the context marker sequence are subjected to encoding processing to obtain a context overall representation; Predicting the context marker sequence and the overall context representation based on the hybrid expert model layer to obtain a text prediction result, and matching the text prediction result in a first preset sensitive word library to obtain a third matching result; If the third matching result is that the text prediction result exists in the first preset sensitive word library, the original complete lyrics information is generated according to the text prediction result.

8. A device for processing lyrics information, characterized in that: include: A complete lyrics information matching module, used to match the original complete lyrics information corresponding to the coded lyrics information from a preset song database; An uncoded lyrics line matching module, used for matching an uncoded lyrics line corresponding to the coded lyrics line from the original complete lyrics information according to the coded lyrics line in the coded lyrics information; An accuracy verification module, used to verify the accuracy of the coded content in the coded lyrics line according to the uncoded lyrics line; The coded content recovery module is used to recover the coded content of the coded lyrics line according to the uncoded lyrics line when it is determined that the coded content of the coded lyrics line has errors.

9. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed by a processor, the method for processing lyrics information described in any one of claims 1-7 is implemented.

10. An electronic device, characterized in that: include: processor; as well as A memory, configured to store executable instructions of the processor; Wherein, the processor is configured to execute the method for processing lyrics information as described in any one of claims 1-7 by executing the executable instructions.

Images