Dear colleagues,

In the context of the Lexhnology ANR project (joint linguistic and NLP discourse structure modelling of legal texts for language pedagogy), started early 2023, we currently have one open position for doctoral candidates with a background in Natural Language Processing or related fields.

# Thesis topics

Interest in the legal field has recently exploded in the NLP community. International evaluation campaigns are proposed on several semantic tasks such as legal information extraction, entailment, rhetorical role recognition, judgement prediction (LegalEval@SemEval2023, COLIEE-2023)... In addition, several conferences and workshops gathering researchers have been recently organised (ASAIL@ICAIL2023, JURISIN@IsAI-2023), showing the growing interest of the NLP community in this specific domain. Numerous datasets are now built and collected (PileOfLaw20, LexGLUE22), allowing the community to create specialised Large Language Models (LLMs) in the legal field (e.g. LegalBERT).

This craze is due to the fact that legal texts have several specific characteristics that make their automatic processing difficult and require specific development: they are both language and domain specific and often longer than the length LLMs can handle.

The role of the PhD student to be recruited will be to: - propose a framework for probing Pretrained Language Models in terms of the captured discourse information - research effective methods to inject discourse knowledge in Transformer-based language models (discourse inspired self-learning tasks or multi-tasks learning or Transformer architecture revision...) - develop an argumentative structure recognition system which will be used in an online platform by legal English users for supporting their reading and understanding tasks

# Project context

The PhD fellowship is offered in the context of the Lexhnology (joint linguistic and NLP discourse structure modelling of legal texts for language pedagogy) project funded by the French National Research Agency (https://lexhnology.hypotheses.org/). Partenaires include CRINI (Nantes Université), LS2N (Nantes Université), ATILF (CNRS & Université de Lorraine), and LAIRDIL (Université de Toulouse).

The successful candidate will join the NLP research group at LS2N lab in Nantes (https://taln-ls2n.github.io/). Nantes is located in the western part of France, crossed by the Loire River, and situated just 50 kilometres away from the Atlantic coast (https://www.levoyageanantes.fr/en/to-see/).

# Requirements

* Master degree (completed or nearly completed) in Computer Sciences, Computational Linguistics, Natural Language Processing, Machine Learning, Data Sciences or a closely related field * Excellent academic records * Practical experience in Machine Learning (esp Deep Learning) methods * Good knowledge of experimental design methodology and statistics * Some level of familiarity with discourse analysis would be a plus * Excellent programming skills (esp. Python) * English (at least B2) and French proficiency both spoken and written * Initiative and ability to work independently and as part of a team

# General information

* Supervisors: Prof. Richard Dufour, Dr Nicolas Hernandez, Dr Laura Monceaux. * Type of Contract : PhD Student contract / Thesis offer * Contract Period : 36 months * Start date of the thesis : 1 September 2023 * Proportion of work : Full time on site * Remuneration : about 2175 € gross monthly (before taxes), partial reimbursement of public transport costs

# Additional information and application

Application deadline : 8 May 2023

For further information and application, contact Nicolas Hernandez (nicolas.hernandez@univ-nantes.fr) AND Laura Monceaux (laura.monceaux@univ-nantes.fr) AND Richard Dufour (richard.dufour@univ-nantes.fr).

Applications should contain all the documents indicated below: a. Free style cover letter outlining the interest for the PhD/ANR project b. Curriculum vitae c. Transcripts of grades from first and second year of master's program (or, if applicable, a document attesting to anticipated success) d. Names and addresses of two references.

Shortlisted applicants will be interviewed online.