|| Master Thesis Seminar at the Signals & Systems Group
of a query refinement method for user preference based information retrieval
of digital music data
Speaker: Erik Zeitler
- Time and Place:
- Tuesday, June 4th, at 13:15
Library, floor 2 at Magistern, Dag Hammarskjölds väg 31, Uppsala
The main task of conventional music retrieval systems is to retrieve
music data which matches the request of a user. The importance of
such systems is expected to increase due to the rapid spread of digital
music data, such as MP3. However, conventional systems are only able
to search a particular song from a database. New methods need to be
implemented that can do complex music data matchings, such as music
retrieval based on user preferences.
In this thesis work a content-based music information retrieval method
that matches songs to user preferences is implemented. The method
constructs a query by encoding a musical preference profile from a
set of rated songs provided by the user. Songs that are similar to
the contents of the highest rated songs in the query are retrieved
from a music collection. Relevance feedback techniques are applied
to further improve the performance of the method. Through evaluation
experiments, the effectiveness of the method is proven.
Conference paper in Pdf.