Identifying Gene Function Descriptions by Probability-based Sentence Selection.

摘要: This paper proposes an approach to the secondary task in TREC Genomics Track. We regard as identification of sentences describing gene functions (i.e., GeneRIFs) and propose a method considering two factors: topicality relevance. The former refers sentence is measured based on location information word frequencies article. latter relevance GeneRIF vocabulary used formalize probabilistic model combining these features. Our evaluated test set 139 MEDLINE abstracts, results demonstrate that (a) function words input could help identify descriptions (b) there peculiar GeneRIFs (c) shows highest predictive power for this particular despite its simplicity. Additionally, we examine some alternative methods comparison with our method.