Legal Natural Language Processing Lab

Instructors: Prof. Matthias Grabmair

Course language: English

2 SWS, 4 ECTS

Session Times: TBA Information Session

Meeting Recording:

https://tum.cloud.panopto.eu/Panopto/Pages/Viewer.aspx?id=98da5f72-9fab-4cee-85d2-b1aa00b22fef

[IMPORTANT]

Following is a short questionnaire meant to pre-assess your background in NLP suffices for the seminar, and to provide us with information on how to rank applicants since we have limited slots available. If you are interested in the seminar and would like to match with us, please fill out the form. Questionnaire Link 

Content Outline

Advances in Natural Language Processing (NLP) technology are increasingly absorbed into support systems for legal practitioners. At the same time, the legal domain, and legal text in particular, move into the focus of mainstream NLP research as a source of challenging problems and opportunities to make a real world impact.

This seminar will offer a curated set of topics in natural language processing of legal text for students to explore and distill into a paper and presentation. The expectation is that seminar contributions should investigate and discuss (1) how legal NLP tasks are similar/different to those explored in mainstream NLP, (2) what technical challenges arise when tackling particular tasks, (3) how NLP methods and models behave on legal data, (4) what non-legal mainstream NLP literature is relevant for the topic, and (5) what implications the obtained results have for practical application and future research. Seminar papers and presentations should critically discuss prior work in depth. In order to do this effectively within the scope of the seminar, students should have prior knowledge in NLP. Some seminar topics may include basic technical work (e.g., testing large language models with prompts, exploring a dataset, comparing model configurations, etc.).

· Representative examples of seminar topics may include:

· Case outcome prediction in different jurisdictions and model architectures

· State of the art in pretraining of large legal language models

· The past, present, and future of legal argument mining

· Automatic compliance checking of legal requirements

· State of the art in legal search engines

· Potentials and Limits of Reasoning Abilities in Large Language Models

· State of the art in non-English legal NLP for selected tasks

Students will conduct a survey of academic research literature around a given topic in natural legal language processing, write a scientific survey about the gained insights, and present it to, and discuss it with, other seminar participants. Grading will be based on the survey document and presentation. Learning Outcomes

After completing this module, students will have gained experience in conducting a literature survey around a given topic in natural legal language processing, writing a survey paper according to academic standards, and present/discuss their findings with peers. In particular, they will have gained experience in:

· Searching and collecting relevant source material for a topic

· Curating the material for inclusion into the paper

· Critically reading, and reflecting on, relevant research literature

· Distilling the insights gained from the survey into a seminar paper narrative

· Writing a scientific seminar paper according to academic standards

· Prepare and conduct a presentation, and discussion, of the findings Requirements

No strict course requirements, but experience in Natural Language Processing is necessary to succeed at the lab. Recommended prior courses are:

· IN2332: Statistical Modeling and Machine Learning

· IN2062: Grundlagen der k√ºnstlichen Intelligenz / Foundations of Artificial Intelligence

· IN2361: Natural Language Processing

· IN2395: Legal Data Science & Informatics

· IN2106: Legal Natural Language Processing Lab (Praktikum)