City Research Online

Modeling Interval Relations for Neural Language models

Kopparti, R. M. and Weyde, T. ORCID: 0000-0001-8028-9905 (2019). Modeling Interval Relations for Neural Language models. Machine Learning for Music Discovery, ICML, Long Beach, June 9-15, 2019, 97,

Abstract

In this study, we explore the use of modellingof pitch intervals and interval relations in pitchwith neural networks. Intervals and their relationsare essential features of music, but in neural net-works, the trend is to use raw data as input andnot to model any higher level aspects of the music.We propose to use Relation Based Patterns (RBP)to integrate intervals (early and mid fusion) andinterval relations (late fusion) into the networkstructure. We observe significant improvementsin pitch prediction for the Essen Folk Song Col-lection for RBP over standard networks, and formixed over unsigned and signed interval represen-tation.

Publication Type: Article
Subjects: Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Departments: School of Mathematics, Computer Science & Engineering > Computer Science
Date Deposited: 09 Sep 2020 09:37
URI: https://openaccess.city.ac.uk/id/eprint/24728
[img]
Preview
Text - Accepted Version
Download (152kB) | Preview

Export

Downloads

Downloads per month over past year

View more statistics

Actions (login required)

Admin Login Admin Login