17-09-2012, 10:16 AM
Using User Models in Music Information Retrieval Systems
usermodeling.pdf (Size: 63.12 KB / Downloads: 37)
ABSTRACT
To make multimedia data easily retrieved, we use
metadata to describe the information, so that search
engines or other information filter tools can effectively
and efficiently locate and retrieve the multimedia content.
Since many features of multimedia content are
perceptual and user-dependent, user modeling is also
necessary for multimedia information retrieval systems,
e.g., music information retrieval systems. Furthermore,
to make the user models sharable, we need standardized
language to describe them. In this paper, an XMLlike
language is proposed to describe the user model for
music information retrieval purposes. We also propose
some paradigms to acquire, deploy and share the user
information to improve current music information
systems. A prototype system, MusicCat, is analyzed
and implemented as a case.
INTRODUCTION
The Internet has become one of the main sources of
multimedia content for entertainment, education, business,
or other purposes. Effective and efficient methods
to retrieve data from Internet are desired, and music is
one of the important media falling in this situation.
There are many websites for music sale, advertisement
and sharing. However, the interfaces for users to find
things are not convenient, since only category-based
browsing and/or text-based searching are supported on
these sites. There has been some research to improve
the interface. For pull applications, content-based
search techniques are studied and some prototype
query-by-humming systems have been developed.
Users can input the main melody of the music via web
browser by uploading pre-recorded humming [3][4][5],
or typing the string to represent the melody contour
[6][7]. There are also some systems [8][9], in which
users can search the sound libraries according to the
acoustic features like sound effect, rhythm, etc. For
push applications, collaborative filtering model is
mostly used to recommend music to users.
USER MODELING LANGUAGE
As mentioned above, user information is very valuable
and needs to be shared in the future to make universal
information retrieval possible. Here we propose an
XML-like language, UMIRL (User Modeling for Information
Retrieval Language), in which different
systems may describe the user in this standard format
to make the user model sharable and reusable. Although
it is designed for music information retrieval
purpose in this paper, it can be extended for general
multimedia information retrieval as well.
CONCLUSIONS AND FUTURE WORK
To make music information retrieval systems more
efficient, both user modeling techniques and descriptions
are important. Those are prerequisites for open
and efficient personalized service. We propose the user
modeling language because there hasn’t been much
attention on the ways to share the valuable user data.
We believe the main issues discussed in this paper will
be the most significant but also the hardiest points in
this research area.
For MusicCat system, it would be better if we integrate
learning techniques to do user modeling and support
interface customization. Some wireless communication
technology can be used to make the system portable.
We can also use sensor-driven interface that can automatically
detect the user context and then play corresponding
music.