Legal Search
Legal research is a key aspect of legal practice, particularly for litigators. The average associate may spend up to 35 % of her time conducting legal research (cf. American Bar Association’s 2013 Legal Technology Survey Report).
According to several discussions with leading law firms, clients are also increasingly unwilling to pay for legal research. This might explain why legal search tools is the “dinosaur” of LegalTech applications and are available in all (industrial) countries at least in some form. The purpose of legal search tools is to make different types of static legal content available for human consumption. Computerized legal research began in the mid-1960s. Through the years, computerized legal search has become more sophisticated, efficient and precise. Previously, legal search depended on determining the right keyword. The major step was the change from keywords to semantics. Semantic search is normally defined “as a kind of data searching technique in which a search query aims to not only find keywords matches [...] to determine the contextual [and intent] meaning” of words (Marios Polycarpou et al., Hybrid Artificial Intelligence Systems, 2014, 540). By understanding and connecting intention and context, semantic search engines are able to understand the different motivation and expectation behind queries. A lawyer may input a query in natural language, and the search engine is still able to make sense of it.Legal search is challenging for many reasons. In order to work, legal search tools require a great deal of manual encoding of the legal knowledge. Ideally, the knowledge-encoding could be done automatically, using some kind of NLP system. In practice, automatization is not possible, yet. Additionally, and this is a fundamental problem, law and judicial decisions are not objective and unambiguous. There might be disagreement about the meaning or the implications of a certain case or word. In general, judicial decisions, for instance, do not contain a fixed set of facts and/or rules which can be enumerated a priori.
4.3