The importance of fillers for text representations of speech transcripts

TL;DR

This paper investigates the role of fillers in spoken language understanding, demonstrating that representing fillers with deep contextualized embeddings enhances modeling spoken language and improves downstream task performance.

Contribution

It introduces a method for representing fillers using deep contextualized embeddings, showing their importance in SLU tasks and downstream applications.

Findings

Improved performance on stance prediction task

Enhanced modeling of spoken language with fillers

Fillers' representations contribute significantly to downstream tasks

Abstract

While being an essential component of spoken language, fillers (e.g."um" or "uh") often remain overlooked in Spoken Language Understanding (SLU) tasks. We explore the possibility of representing them with deep contextualised embeddings, showing improvements on modelling spoken language and two downstream tasks - predicting a speaker's stance and expressed confidence.