| Summary: | 碩士 === 國立臺灣師範大學 === 圖書資訊學研究所 === 100 === Music can let people express their emotions (Juslin &; Laukka, 2004) and also their feelings. Under the rapid rise of Internet and online music, listening to online music has become one of everyone's major leisure activities. How to help users find the music which meet their music information needs is an important issue. Most of the online music retrieval systems use the traditional musical bibliographic metadata (such as artist, song title, genre, etc.) to describe and classify music collection. However, such classification doesn't meet users' unknown music information needs. This study suggests that in addition to traditional music attributes, music retrieval system should also provide a variety of user-centered music attributes.
This study collects 500 online music queries of unknown music information needs from the the KKBOX and PTT music forum. By doing content analysis, we found 48 attributes from the music queries which can be summarized as 6 categories. It shows that users use a lot of subjective descriptions with music cognition, feelings, ideas and contexts in addition to objective descriptions to help describe the unknown music information needs.
In addition, this study invited a total of 24 users to join the music describing experiment to collect their music descriptions. The experiment includes four different tasks to explore all potential music attributes from a variety of contexts. All information gathered from tasks will be analyzed through Content analysis. The results show that there are a total of 74 attributes and can be summarized as 7 categories. We can see different phases of music characteristics. Furthermore, the results show that visual description can help music recognition and description.
Overall, the users’ cognition and description of music and music information needs are extensive and diverse. The descriptions are full of association and high connection with users. Moreover, contexts of music listening and uses have great impact on users' music description.
Based on the findings, this study offer the following recommendations: music retrieval systems should provide complete and diverse music attributes, allowing users to find the music in accordance with the specific context, mood or purpose of use; for better music recommendation service, it should provide multiple attributes choices, allowing users to cross search the music; for better music discovery service, it is recommended to provide attributes like "music images" and "association objects" to increase the opportunities of music encountering. At last, all recommendations are to help users have better music discovery experiences.
|
|---|
