![]() More specifically, given a segment s i, i.e., ( a i, i, t i ), of a piece of music M, the lyrics density of s is defined. Finally, lyrics density of each sentence is calculated. A sentence is split into words by blank spaces. It is considered that contents between two adjacent time tags belong to one single lyrics sentence. Time tags in a lyrics file are used to divide the file into lyrics sentences. Second, the cleaned lyrics file is parsed to extract the number of words as well as the time information of each sentence, i.e., the beginning time and the end time. This kind of ‘dirty’ contents may result in an extra number of words in a sentence so that the value of lyrics density will increase mistakenly. Raw lyrics files may contain varieties of ‘dirty’ contents, e.g., advertisement information, because parts of the lyrics that online music websites use are created by audiences. 2) is cleaned to increase the accuracy of lyrics density estimation. ![]() s i = ( a i, i, t i ) is the i th segment of, n the number of segments of M, a i the audio data of s i, i the related lyrics of s i, t i the length of s i, 0 ≤ i < n, n, i ∈ N.
0 Comments
Leave a Reply. |