Information retrieval on MP3-File Tag ID3

Abstract

Searching song based on text data was easily done, but it wasn’t happen if the searching process refers to file info on ID3. Almost all of web-based retrieval only relies on data from database or filename. It causes difficulties in searching audio files on the internet. This research implements an information retrieval system that can search on the ID3 of MP3 files and evaluates the system performance on returning relevant answers. This research uses 182 MP3 file consists of various sizes and genre Indonesian songs as data test and 5 queries. On this research both lyrics and queries have been stemmed using Kamus Besar Bahasa Indonesia (KBBI) stemmer. The tf.idf weighting and sum of tf.idf are used for returning the result. The sum of tf.idf is based-on standard tf.idf that sum of each keyword on the query. Analysis of the results is conducted by several k-top documents. The result shows that information retrieval on the ID3 MP3 files can be done and the result using stemmer is better than that of without stemmer. MAP value where k = 10 with stemmer is 0.27, whereas without using stemmer is 0.25. For the next k, where k = 15 with stemmer the MAP value is 0.39, while without using stemmer 0.19, last with k = 20, the value of MAP with using stemmer is 0.62, while without using stemmer is 0.15 . The MAP values show that using stemmer is better than the MAP without stemmer, it’s because the use of stemmer can reduce the term variation in order to increase the recall value.