Ranking Case Law by Context Dimensions Using Fuzzy Fingerprints

The goal of this paper is to retrieve relevant legal workplace discrimination cases, written in legal language, with regards to a given user story, written in colloquial language. The approach is to use a combination of semantic, lexical and context dimensions such as the grounds and expressions of discrimination. To further enhance the performance of the system, we will match the fingerprints based on context of the legal cases to the particular user story. Furthermore, we combine all three dimensions trying to overcome the weaknesses of the single approaches. Lexical retrievers, for instance, deliver results that are easy to interpret, however they provide only results for exact-word-matches, which is a problem that is being solved by using semantic similarity. Semantic similarity matches meanings of words and document and does not look at the exact word-match. However, as semantics are numerical representations of words in a higher space, results might be hard to interpret. Furthermore, by using context dimensions we can ensure to give a high emphasize on the grounds and expressions of discrimination which play a critical role in this particular application. The retrieval of legal documents such as case law helps to proactively work against workplace discrimination. In this paper we focus discrimination cases on the LGBTQ+ community. This is because this group faces particular challenges such as high numbers of underreporting, data scarcity, making the problem even more complex.