Towards combined semantic and lexical scores based on a new representation of textual data to extract experimental data from scientific publications.