Legal documents employ a domain specific language that requires ad hoc NLP tools to process and extract information from them. In this projecct, UCA Datalab collaborated with Quantum Analytics in an NLP project for Lefébvre el Derecho S.A., one of the main providers of information services for legal bureaus. The main goals of the project were to develop a custom NER model to structure the names and legal roles of plaintiff and defendant in a legal court ruling. UCA Datalab also developed an information retrieval system that adapted to the user’s notion of similarity dynamically, by training a siamese network that performed a fine tuning over the document embeddings.