The thesis presents an opinion mining system for song lyrics, which can fetch objects of interest and opinion words about them. Finally, opinion mining result is analyzed in terms of time information and musical genre. In the process of constructing the system, many previous works are reviewed and some of them are applied to the thesis and different methods are compared for reaching a best solution (e.g. explore how to fetch objects f interest). As well, the evaluation of the system has been done by running experiments with a collection of song lyrics containing hundreds of documents. The result from the system is compared with manual identification. The evaluation result shows that the system basically can present topics of one song lyrics and opinion words about them. Finally, opinion mining result from a collection of song lyrics can be analyzed and some interesting things are presented, e.g. fetching most common topics, presenting the number of polarity words for each musical type or different year, opinion change on some common topics as time changes. Besides, we develop a program in Java for collecting song lyrics on Internet from one website. The program can help us collect thousands of song lyrics and search information of song publishing year or musical genre on Wikipedia.org. The work in opinion mining for song lyrics is few at present. The thesis finishes an exploration in the subject and the exploration is valuable and useful for future wok.
| Identifer | oai:union.ndltd.org:UPSALLA1/oai:DiVA.org:ntnu-11936 |
| Date | January 2010 |
| Creators | Shu, Hanjie |
| Publisher | Norges teknisk-naturvitenskapelige universitet, Institutt for datateknikk og informasjonsvitenskap, Institutt for datateknikk og informasjonsvitenskap |
| Source Sets | DiVA Archive at Upsalla University |
| Language | English |
| Detected Language | English |
| Type | Student thesis, info:eu-repo/semantics/bachelorThesis, text |
| Format | application/pdf |
| Rights | info:eu-repo/semantics/openAccess |
Page generated in 0.002 seconds