Open Data Platform for Knowledge Access in Plant Health Domain : VESPA Mining

Important data are locked in ancient literature. It would be uneconomic to produce these data again and today or to extract them without the help of text mining technologies. Vespa is a text mining project whose aim is to extract data on pest and crops interactions, to model and predict attacks on crops, and to reduce the use of pesticides. A few attempts proposed an agricultural information access. We can find systems with sociosemantic approach (Turenne & Barbier, 2004), or systems with thesaurus-oriented approach like (Milne et al, 2006; Bartol, 2009; Vakkari, 2010; Laporte et al, 2012). But these systems are not driven by concrete usages like the Vespa Mining platform. Another originality of our work is to parse documents with a dependency of the document architecture (sections: chapters, subchapters...). Lots of works has been done to retrieve named entities like genes, persons, organization... in texts with a good efficiency (Riloff, 1996; Roth & Yih, 2002; Carpenter, 2007; Surdeanu et al, 2011; Krieger et al, 2014). Some state of the art tools are even available (http_Lingpipe, 2015; http_SNER, 2015) but in these approaches, entities are searched in a sentence and relations are often searched in a same sentence. Our approach is different, we search relations in a same section. Our approach (Turenne & Phan, 2015) is based on dictionary matching and on rules defined by users using a tool called Unitex (htt_Unitex, 2015) to detect entities and document design together with cooccurrence analysis to detect relation. High efficiency relation extraction reduces time of browsing in documents to identify relevant information.