Automatic discovery and resolution of protein abbreviations from full-text scientific papers: A light-weight approach towards data extraction from unstructured biological sources

We propose a methodology for discovering and resolving a wide range of protein name abbreviations from the full-text versions of scientific articles, as implemented in our PRAISED framework. Three processing steps lie at the core of our approach: an abbreviation identifi- cation phase, carried out via largely domain-independent metrics based on lexical clues and exclusion rules, whose purpose is to identify all pos- sible abbreviations within a scientific text; an abbreviation resolution phase, which takes into account a number of syntactical and semantic criteria and corresponding optimization techniques, in order to match an abbreviation with its potential explanation; and a dictionary-based pro- tein name identification, which is meant to eventually sort out those ab- breviations actually belonging to the biological domain. We have tested our implementation against the well-known Medstract Gold Standard Corpus and a relevant subset of real scientific papers extracted from the PubMed database, obtaining significant results in terms of recall, pre- cision and overall correctness. In comparison to other methods, our ap- proach retains its effectiveness without compromising performance, while addressing the complexity of full-text papers instead of the simpler ab- stracts more generally used. At the same time, computational overhead is kept to a minimum and its light-weight approach further enhances customization and scalability.