DEVELOPMENT OF A TEXT SEARCH ENGINE FOR MEDICINAL CHEMISTRY PATENTS

Over the last decades, the size of patent collections has strongly increased. Thus, in 2009, it was estimated that there are globally about 50 millions patents (Bonino et al., 2010) with about 15-20 millions related to medicinal chemistry, which represent a corpus of knowledge comparable to the content of MEDLINE. These collections represent an important and high-quality source of knowledge. However, while the past years have seen the development of a wealth of search engines and text mining instruments to navigate the bibliome – a term coined by (Grivell, 2002) to refer to the post-omics biomedical literature – with applications such as EBIMed, EAGLi, GoPubMed and Twease to cite only a few of them, text-mining applications dedicated to patents of the biomedical field remain rare. Recently, the development of such specialized patent retrieval engines has benefited from the effort of dynamic communities of researchers, encouraged by the emergence of several evaluation campaigns, such as the Text REtrieval Conferences (TREC), one of the most popular competitions to evaluate and compare search engines. Prestigious universities, but also corporate research centres have regularly participated to these competitions. Lately, a task of information retrieval dedicated to patent search for chemistry, called TREC-Chem