New team member: Jeremy Auguste
Jeremy Auguste is joining the ENP-China project as a postdoctoral researcher in Natural Language Processing. He completed his PhD in Computer Science at the Laboratoire d’Informatique et Systèmes (LIS) in 2020. Before his PhD, he obtained in 2016 a master’s degree in Fundamental Computer Science at Aix-Marseille Université, with a specialization in Natural Language Processing and Machine Learning.
Jeremy’s PhD was carried out in the context of the DATCHA ANR project, in which he focused his research on finding ways to do conversational discourse analysis on huge quantities of raw online chats while having very few discourse annotations in the data. He proposed several methods related to the representation learning field: trying to use a discourse related task (customer satisfaction prediction) for which data is available in order to indirectly extract discourse related features; learning specialized distributed sentence representations that explicitly take into account speaker interactions (skip-act vectors); using these sentence representations in order to learn discourse parsers from very few annotations. His work was published in several conferences and workshops:
ICASSP, LREC, IWSDS, TALN, RepEval.
After his PhD, he was a postdoctoral researcher in the LIS, on a contract with
Orange Labs (Lannion), while also taking part in the ARCHIVAL ANR project.
During this postdoc, he focused his research on reading comprehension for
question answering by adapting question-answering models to new domains (texts on archaeology and history, humanities journals on workers’ self-management). He proposed a method that allows the automatic production of question-answering corpora on new texts by using transformer-based sequence-to-sequence models and semantic frames in order to generate questions for any kind of text.
Cite this blog post
Christian Henriot (2021, March 14). New team member: Jeremy Auguste. Elites, Networks and Power in modern China. Retrieved February 29, 2024, from https://doi.org/10.58079/o8lm