Gene mention normalization in full texts using GNAT and LINNAEUS

Goran Nenadić, Illés Solt, Martin Gerner, Philippe Thomas, Casey Bergman, Ulf Leser, Jörg HakenbergResearch 2010

Abstract

Gene mention normalization (GN) refers to the automated mapping of gene names to a unique identifier, such as an NCBI Entrez Gene ID. Such knowledge helps in indexing and retrieval, linkage to additional information (such as sequences), database curation, and data integration. We present here an ensemble system encompassing LINNAEUS for recognizing organism names and GNAT for recognition and normalization of gene mentions, taking into account the species information provided by LINNAEUS. Candidate identifiers are filtered through a series of steps that take the local context of a given mention into account. On the BioCreative III high-quality training data, our system achieves TAP-5 and TAP-20 scores of 0.36 and 0.41, respectively. On the evaluation set of 50 documents that were provided to participants, we achieve scores of 0.16 and 0.20 for TAP-5 and TAP-20, respectively. Our analysis of the evaluation results suggests that the lower scores primarily are due to significant differences in species composition, and partly due to the method for selecting the evaluation data.
BibTeX
@inproceedings{nenadic2010gene,
  author = {Goran Nenadić and Illés Solt and Martin Gerner and Philippe Thomas and Casey Bergman and Ulf Leser and Jörg Hakenberg},
  title = {{Gene mention normalization in full texts using GNAT and LINNAEUS}},
  booktitle = {Research},
  year = {2010},
}

Loading related content...

Figure