Anyone who has just come up with a new invention and wants be granted a valid patent has to ensure that the invention is really novel in all aspects, in other words, that it has not been previously patented by somebody else or has been described in a scientific paper. To ensure this, the patent applicant has to perform (or let somebody else perform) a so-called 'prior art search': a search for all documents that are relevant to the invention to see if he is infringing on somebody else's idea. These searches are mainly done using keyword-based search (i.e. typing in separate words like in Google) and and can take a very long time (days or even weeks) because you do not want to miss that one, very important document... . In the TM4IP project we want to introduce a new search engine for patent search which enables the searcher to interactively find the relevant documents. We combine Information Retrieval techniques and linguistic knowledge to build a hybrid parser aimed at technical texts and a search engine that uses linguistic knowledge to find the relevant patents.