Categorization of musicology questions from community-based Q&A site using latent dirichlet allocation

Fu, Hengyi

Permalink

https://hdl.handle.net/2142/96750 Copy

Description

Title

Categorization of musicology questions from community-based Q&A site using latent dirichlet allocation

Author(s)

Fu, Hengyi

Issue Date

2017

Keyword(s)

• Music information retrieval
• Document representation
• Topic models
• Latent dirichlet allocation
• Social Q&A site

Abstract

Question & Answer websites, such as Yahoo! Answer and Stack Exchange, are widely used to look for and exchange information in different knowledge domains. Question-answer pairs extracted from social Q&A sites can be considered as query and returned results in an information retrieval system, and descriptive access points are the intermediary to connect the information needs and information. However, information seekers sometimes experienced difficulty in precisely describing access points, and relied on system-suggested access points to search information. This study is the first attempt to perform automatic topic modeling (Latent Dirichlet Allocation) based on the text of questions and answers related to a wide range of musical topics. The analyses confirm that LDA is a very effective tool to classify textual, unstructured, music-related data and the identified topics and question types are similar to expert cataloging based on the same set of data.

Publisher

iSchools

Series/Report Name or Number

iConference 2017 Proceedings

Type of Resource

text

Language

en

Permalink

http://hdl.handle.net/2142/96750

Copyright and License Information

Copyright 2017 Hengyi Fu

Owning Collections