Contract management is a sensitive and crucial task for companies. Contracts and other legal documents have to be fully searched and understood to define the obligations of the companies towards other actors in the company business. Hence, lawyers are often obliged to spend a tremendous amount of time on tedious and repetitive tasks to exploit the data at their disposal. To address these issues, Hyperlex offers a software solution based on Natural Language Processing (NLP).
Relation Extraction is an active area in NLP research which aims to extract and classify the relationship between two entities. Legal documents contain a lot of such entities and relationships which are specific to the judicial domain and provide capital information about obligation definitions. This master thesis aims to propose novel approaches to deal with these issues and improve Hyperlex's relation extraction and classification tools.