# A case study on using speech-to-translation alignments for language documentation

## TL;DR

This paper explores how speech-to-translation alignments can improve transcription quality in low-resource languages, aiding language documentation and speech recognition training, demonstrated through a small-scale case study.

## Contribution

It introduces a method to leverage speech-to-translation alignments for enhancing crowdsourced transcriptions and proposes a simple phonetic string averaging technique for better transcription quality.

## Key findings

- Alignments improve transcription accuracy
- Phonetically aware string averaging yields higher quality transcriptions
- Potential benefits for speech recognition in low-resource languages

## Abstract

For many low-resource or endangered languages, spoken language resources are more likely to be annotated with translations than with transcriptions. Recent work exploits such annotations to produce speech-to-translation alignments, without access to any text transcriptions. We investigate whether providing such information can aid in producing better (mismatched) crowdsourced transcriptions, which in turn could be valuable for training speech recognition systems, and show that they can indeed be beneficial through a small-scale case study as a proof-of-concept. We also present a simple phonetically aware string averaging technique that produces transcriptions of higher quality.