English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Chord recognition systems use temporal models to post-process frame-wisechord preditions from acoustic models. Traditionally, first-order models suchas Hidden Markov Models were used for this task, with recent works suggestingto apply Recurrent Neural Networks instead. Due to their ability to learnlonger-term dependencies, these models are supposed to learn and to applymusical knowledge, instead of just smoothing the output of the acoustic model.In this paper, we argue that learning complex temporal models at the level ofaudio frames is futile on principle, and that non-Markovian models do notperform better than their first-order counterparts. We support our argumentthrough three experiments on the McGill Billboard dataset. The first two show1) that when learning complex temporal models at the frame level, improvementsin chord sequence modelling are marginal; and 2) that these improvements do nottranslate when applied within a full chord recognition system. The third, stillrather preliminary experiment gives first indications that the use of complexsequential models for chord prediction at higher temporal levels might be morepromising.

On the Futility of Learning Complex Frame-Level Language Models for Chord Recognition