CN-121997916-A - Method and device for identifying singing song, electronic equipment, storage medium and product

CN121997916ACN 121997916 ACN121997916 ACN 121997916ACN-121997916-A

Abstract

The invention relates to the technical field of information retrieval and discloses a method, a device, electronic equipment, a storage medium and a product for identifying a singing song, wherein the method comprises the steps of obtaining a first song lyrics of a song to be identified and a second song lyrics of a target song; and determining a sung identification result according to the similarity of the chorus, wherein the sung identification result indicates whether the song to be identified belongs to the sung song of the target song or not. The invention can effectively reduce the misjudgment rate of the eversion songs and improve the accuracy of judging the eversion songs.

Inventors

LI CHENGKAI
YU JIANQIANG
ZHOU SIHANG
CHEN JINHAI

Assignees

杭州网易云音乐科技有限公司

Dates

Publication Date: 20260508
Application Date: 20251229

Claims (10)

1. A method for identifying a singing song, the method comprising: acquiring a first song lyrics of a song to be identified and a second song lyrics of a target song; text matching is carried out on the first auxiliary song lyrics and the second auxiliary song lyrics, so that auxiliary song similarity between the song to be identified and the target song is obtained; And determining a sung identification result according to the similarity of the verses, wherein the sung identification result represents whether the song to be identified belongs to the sung song of the target song.
2. The method of claim 1, wherein the text matching the first and second sub-song lyrics to obtain a sub-song similarity between the song to be identified and the target song comprises: dividing the first chorus lyrics into a plurality of continuous clause segments to obtain a first clause segment sequence; dividing the second sub-song lyrics into a plurality of continuous clause segments to obtain a second clause segment sequence; Respectively obtaining the similarity between each first clause segment sub-sequence in the first clause segment sequence and the second clause segment sequence, wherein each first clause segment sub-sequence comprises a plurality of continuous clause segments in the first chorus lyrics; And taking the highest acquired similarity as the chorus similarity.
3. The method according to claim 2, wherein the separately obtaining the similarity between each first clause segment sub-sequence in the first clause segment sequence and the second clause segment sequence comprises: Traversing each first clause segment sub-sequence in the first clause segment sequence in turn through a sliding window based on a sliding window algorithm; And acquiring the similarity between the first sentence fragment subsequence currently traversed by the sliding window and the second sentence fragment sequence by adopting a preset text similarity algorithm.
4. The method of claim 1, wherein the determining a chorus recognition result from the chorus similarity comprises: and determining the song to be identified as the singing song of the target song in response to the similarity of the chorus being greater than or equal to a first preset threshold value, and obtaining the singing identification result.
5. The method of claim 1, wherein the determining a chorus recognition result from the chorus similarity comprises: and in response to the chorus similarity being smaller than a first preset threshold and larger than or equal to a second preset threshold, inputting the first chorus lyrics and the second chorus lyrics into a preset large model to perform a chorus analysis, and obtaining a chorus recognition result.
6. The method of claim 1, wherein the obtaining the first song lyrics of the song to be identified and the second song lyrics of the target song comprises: Identifying the chorus position of the song to be identified from the audio of the song to be identified and identifying the chorus position of the target song from the audio of the target song by an audio-based chorus extraction algorithm; Extracting the first chorus lyrics according to the chorus position of the song to be identified; And extracting the second chorus lyrics according to the chorus position of the target song.
7. A singing song recognition apparatus, the apparatus comprising: the acquisition module is used for acquiring the first song lyrics of the song to be identified and the second song lyrics of the target song; the matching module is used for performing text matching on the first and second chorus lyrics to obtain the chorus similarity between the song to be identified and the target song; And the identification module is used for determining a sung identification result according to the chorus similarity, wherein the sung identification result represents whether the song to be identified belongs to the sung song of the target song.
8. An electronic device, comprising: A memory and a processor, said memory and said processor being communicatively coupled to each other, said memory having stored therein computer instructions, said processor executing said computer instructions to perform the method of identifying a singing song as claimed in any one of claims 1 to 6.
9. A computer-readable storage medium having stored thereon computer instructions for causing a computer to perform the method of identifying a music-on-music-turned as claimed in any one of claims 1 to 6.
10. A computer program product comprising computer instructions for causing a computer to perform the method of identifying a singing song as claimed in any one of claims 1 to 6.

Description

Method and device for identifying singing song, electronic equipment, storage medium and product Technical Field The invention relates to the technical field of information retrieval, in particular to a singing song identification method, a singing song identification device, electronic equipment, a storage medium and a singing song identification product. Background In the related art, song overturn recognition is mainly performed by judging whether the lyrics of two songs mutually contain the same lyrics. In this way, when the same common sentence appears in the lyrics of two songs, the song is easily judged to be a music turning song, but the music does not actually belong to a music turning song, and when the music turning song only carries out fine change on the core sentence in the lyrics of the original song, but the meaning of the actual lyrics expression is not changed, the song is easily judged to be a music turning song, but the music turning song actually belongs to the music turning song. Therefore, the song overturn identification mode of the related technology has serious misjudgment, and the song overturn identification accuracy is lower. Disclosure of Invention The invention provides a method, a device, electronic equipment, a storage medium and a product for identifying a singing song, which are used for solving the problem that song singing identification accuracy is low in related technologies. In a first aspect, the present invention provides a method for identifying a singing song, the method comprising: acquiring a first song lyrics of a song to be identified and a second song lyrics of a target song; text matching is carried out on the first auxiliary song lyrics and the second auxiliary song lyrics, so that auxiliary song similarity between the song to be identified and the target song is obtained; And determining a sung identification result according to the similarity of the verses, wherein the sung identification result represents whether the song to be identified belongs to the sung song of the target song. In an optional implementation manner, the text matching of the first and second sub-song lyrics to obtain the sub-song similarity between the song to be identified and the target song includes: dividing the first chorus lyrics into a plurality of continuous clause segments to obtain a first clause segment sequence; dividing the second sub-song lyrics into a plurality of continuous clause segments to obtain a second clause segment sequence; Respectively obtaining the similarity between each first clause segment sub-sequence in the first clause segment sequence and the second clause segment sequence, wherein each first clause segment sub-sequence comprises a plurality of continuous clause segments in the first chorus lyrics; And taking the highest acquired similarity as the chorus similarity. In an alternative embodiment, the separately obtaining the similarity between each first clause segment sub-sequence in the first clause segment sequence and the second clause segment sequence includes: Traversing each first clause segment sub-sequence in the first clause segment sequence in turn through a sliding window based on a sliding window algorithm; And acquiring the similarity between the first sentence fragment subsequence currently traversed by the sliding window and the second sentence fragment sequence by adopting a preset text similarity algorithm. In an alternative embodiment, the text similarity algorithm includes a string edit distance algorithm, and the obtaining, by using a preset text similarity algorithm, a similarity between a first clause segment sub-sequence and the second clause segment sequence currently traversed by the sliding window includes: Acquiring the editing distance between the character string of the first sentence fragment subsequence currently traversed by the sliding window and the character string of the second sentence fragment sequence by adopting the character string editing distance algorithm; and obtaining the similarity according to the editing distance, the character string length of the first clause segment sub-sequence currently traversed and the character string length of the second clause segment sequence. In an optional implementation manner, the determining the chorus recognition result according to the chorus similarity includes: and determining the song to be identified as the singing song of the target song in response to the similarity of the chorus being greater than or equal to a first preset threshold value, and obtaining the singing identification result. In an optional implementation manner, the determining the chorus recognition result according to the chorus similarity includes: and in response to the chorus similarity being smaller than a first preset threshold and larger than or equal to a second preset threshold, inputting the first chorus lyrics and the second chorus lyrics into a preset large model to perform a chorus an