Creation and Enrichment of a Terminological Knowledge Graph in the Legal Domain

Domain-specific terminologies are of great use in a number of contexts, such as information retrieval from text documents or supporting humans in translation tasks. However, automated terminology extraction tools usually render plain lists with no additional information (hierarchical relations, definitions or examples of use, amongst others). The output of these tools is very often offered in non-open formats, hampering their reuse and interoperability. Moreover, terminology management tools demand a lot of manual work to curate and enrich the resources and they do not support the representation of terminological relations beyond broader/narrower. The contributions of this Thesis mitigate these problems by automating the creation of rich terminologies from plain text documents, by establishing links to external resources, and by adopting the W3C standards for the Semantic Web. The proposed method comprises six tasks: refinement, disambiguation, enrichment, relation validation, relation extraction and RDF conversion. We have applied this methodology to two different legal corpora, i.e., contracts and collective agreements. The result of this methodology will be a Terminological Knowledge Graph that can be exploited by different Natural Language Processing applications.