Party Identification of Legal Documents 
Advised by: Amal Shehan Perera, Nisansa de Silva, Gathika Ratnayaka
University of Moratuwa
Law and order is the core guideline in a society that is predominantly responsible for maintaining the peace and conduct among citizens. It in essence keeps society running for without law there would be chaos and it would be survival of the fittest and every man for himself. Not an ideal lifestyle for most part. Therefore, it is obvious that any advancement made to the field of law is in fact an advancement done towards the betterment of the society. This research is an effort made to contribute to a legal system that would be well capable of extracting information from court cases and analyzing them, providing the users with easy and meaningful access to the information residing in them. This research particularly looks at identifying the legal entities involved in a given legal case.
Case law is a form of law that is created by the courts in the process of passing judgement for a case. With the existence of case law, interpretations of the laws done in previously decided cases can be used as precedents to support arguments in an ongoing trial. Therefore, it is expected of the lawyers and other legal officials to be knowledgeable on cases that are similar to the one in the hearing. With a setting as such, arises the necessity of an automated system that can extract information out of legal documents with a Natural Language Understanding(NLU) approach. The tediousness associated with browsing through the excessive amount of documents against the limited time available to examine them further motivates for such a system. A legal opinion is a form of legal documentation that can immensely aid as a resource in building such a system since these documents contain legal conclusions and analysis done by lawyers regarding a legal transaction. And when observing opinion texts, it can be seen that the arguments and the counter-arguments presented there often stem out from the parties involved in the case. Therefore, we believe that legal Party extraction stands out with high significance in the process of extracting meaningful information out of legal opinion texts. Nevertheless, we face many challenges in the process of advocating Natural Language Processing(NLP) into working with legal documents. Learning the Rulebook: Challenges Facing NLP in the Legal Contexts points out three roadblocks to incorporate NLP in legal context, as: 1) the unique hierarchical structure of outcomes, 2) the linguistic quirk of legal adversarial and 3) the challenge of using acontextually trained embeddings.
In addition to these more generic issues found with using NLP for the legal domain, the ambiguous nature of the legal parties also makes the party extraction complicated. That is the fact that a mere person's appearance in the text does not qualify the said entity to be a legal party; for example, a witness in a case is not considered to be a party involved in that case. Additionally, in opinion texts title terms like petitioner, defendant appear before an entity name. Nevertheless, it cannot be identified with full confidence that an entity having the said title belongs to the same party since they could have acted according to the title in a previous case but not in the case at discussion.
Keywords: Natural Language Processing | Machine Learning / Deep Learning | Law | Legal Party Identification | Legal Domain |