Lexsnap
Client: Tetiana Bersheda <tetiana@lexsnap.com>
Perhaps: LawBot
proposal
Law tech startup Lexsnap is building a web application where users can instantly access legal advice. This is powered primarily through question-matching, where the most relevant question-answer pair is shown for a particular query. The underlying knowledge-base consists of 500 manually created question-answer pairs focused on Family Law. Your task will be to utilise NLP techniques to explore ways of automatically adding new areas of law into the knowledge base. This will consist of extracting snippets of information from UK legal sources, converting them to a question-answer format and automatically simplifying the language.
response
This is certainly an area that would be appropriate for one of these projects. I’m fairly familiar with this kind of system, having been involved in the early days of Transversal, whose original product was also based on question-answer pair matching.
We like to be able to offer students the chance to design an end-to-end solution of some kind, in part to gain experience of problem analysis, but also so that the each member of the six-person team can address different aspects of the problem.
We could perhaps suggest that they create something along the lines of your current service (including construction of the question/answer bank, and thinking about the most appropriate metrics for approximate matching of new questions), but maybe in a different application domain?
Alternatively, they could experiment with an approach that works in the legal domain, but disguises or replaces the basic question-answer mechanic, for example using an automated chatbot interaction.
I’d be happy to take a look in more detail at your existing service, if you could give me permission to do so without legal constraint (I did take an initial look, but I noted that the terms and conditions say that I cannot access the site without agreeing to the terms and conditions - seems to be a catch 22, as I had to access the site to see them, but couldn’t agree to them without seeing them :-)
