Lecture | Anthropological & Descriptive Discussion Group
Accelerating transcription of fieldwork data using machine learning
- Monday 16 December 2019
2311 BD Leiden
This talk will introduce Elpis, a user-friendly toolkit designed to help linguists transcribe fieldwork recordings faster. Elpis facilitates the transcription process by making it easy to train and apply a machine-learning speech recognition model on linguistic fieldwork data. We will describe the history of the Elpis project and its development, and provide some example scenarios in which Elpis has been applied so far. We will also introduce the current user interface and describe at a very high level how the underlying machine learning technology works. We will also describe what other possibilities language technology offers for accelerating the annotation and analysis process in general.
About the speaker
Daan van Esch is a Technical Program Manager with Google AI in Mountain View, CA, working on the Languages & Linguistics team that's tasked with bringing language technologies such as smart keyboards and voice dictation systems to more languages around the world. Prior to joining Google seven years ago, Daan completed a BA and an MA in Chinese Studies at Leiden, with a focus on linguistics. His MA research focused on using large data sets to study regional linguistic variation in China. Daan also works with the Australian Centre of Excellence for the Dynamics of Language (CoEDL) on applying AI to assist linguists in the process of language documentation.