Skip to Main content Skip to Navigation
Preprints, Working Papers, ...

La transcription du linguiste au miroir de l’intelligence artificielle : réflexions à partir de la transcription phonémique automatique

Abstract : Automatic speech recognition systems now achieve high levels of accuracy with relatively small amounts of training data: on the order two to three hours of transcribed speech, instead of tens of hours for previous tools. Beyond the practical usefulness of these technological advances for linguistic documentation tasks, use of automatic transcription also yields some linguistic insights. Acoustic models are built on the basis of the linguist’s transcriptions, and thus encapsulate linguistic hypotheses and assumptions. To what extent can acoustic models be examined in turn by the linguist? What can we learn from this renewed confrontation with the acoustic signal? The present study is based on examples from the Native language (Sino-Tibetan family) to illustrate how error analysis allows a renewed confrontation with the data. Among other benefits, error analysis allows for a renewed exploration of phonetic detail: examining the output of phonemic transcription software compared with spectrographic and aural evidence. Some reflections on experiments of automatic transcription of the Tsuut'ina language (Dene family) are also presented.
Document type :
Preprints, Working Papers, ...
Complete list of metadatas

Cited literature [63 references]  Display  Hide  Download

https://halshs.archives-ouvertes.fr/halshs-02881731
Contributor : Alexis Michaud <>
Submitted on : Friday, June 26, 2020 - 8:36:15 AM
Last modification on : Thursday, July 2, 2020 - 3:46:37 AM

Files

TranscriptionAutomatique_BullS...
Files produced by the author(s)

Licence


Distributed under a Creative Commons Attribution - NonCommercial - ShareAlike 4.0 International License

Identifiers

  • HAL Id : halshs-02881731, version 1

Citation

Alexis Michaud, Oliver Adams, Christopher Cox, Séverine Guillaume, Guillaume Wisniewski, et al.. La transcription du linguiste au miroir de l’intelligence artificielle : réflexions à partir de la transcription phonémique automatique. 2020. ⟨halshs-02881731⟩

Share

Metrics

Record views

26

Files downloads

33