Honors College Thesis
 

Predicting music emotion with social media discourse

Public Deposited

Downloadable Content

Download PDF
https://ir.library.oregonstate.edu/concern/honors_college_theses/tt44pv32b

Descriptions

Attribute NameValues
Creator
Abstract
  • Predicting the average affect of a piece of music is a task which has been of recent interest in the field of music information retrieval. We investigate the use of sentiment analysis on online social media conversations to predict a song’s valence and arousal. Using four music emotion datasets - DEAM, AMG1608, Deezer, and PmEmo, we create a corpus of social media commentary surrounding the songs contained in these datasets by extracting comments from YouTube, Twitter, and Reddit. Two learning approaches are compared − one bag-of-words model using dictionaries of affective terms to extract emotive features, and a DistilBERT transformer model fine-tuned on our social media discourse to perform direct comment-level valence and arousal prediction. We find that transformer models are better suited to the task of predicting music emotion directly from social media conversations.
Resource Type
Date Issued
Degree Level
Degree Name
Degree Field
Degree Grantor
Commencement Year
Advisor
Committee Member
Non-Academic Affiliation
Rights Statement
Related Items
Publisher
Peer Reviewed
Language
Embargo reason
  • Ongoing Research
Embargo date range
  • 2022-08-24 to 2023-09-20

Relationships

Parents:

This work has no parents.

In Collection:

Items