Centering: a framework for modeling the local coherence of discourse
Computational Linguistics
Medstract: creating large-scale information servers for biomedical libraries
BioMed '02 Proceedings of the ACL-02 workshop on Natural language processing in the biomedical domain - Volume 3
RelEx---Relation extraction using dependency parse trees
Bioinformatics
Semi-supervised anaphora resolution in biomedical texts
BioNLP '06 Proceedings of the Workshop on Linking Natural Language Processing and Biology: Towards Deeper Biological Literature Analysis
Quantitative data on referring expressions in biomedical abstracts
BioNLP '07 Proceedings of the Workshop on BioNLP 2007: Biological, Translational, and Clinical Language Processing
Anaphora resolution for biomedical literature by exploiting multiple resources
IJCNLP'05 Proceedings of the Second international joint conference on Natural Language Processing
| Hi-index | 0.00 |
Drug-drug interactions are frequently reported in biomedical literature and Information Extraction (IE) techniques have been devised as a useful instrument for managing this knowledge. Nevertheless, IE at the sentence level has a limited effect because there are frequent references to previous entities in the discourse, a phenomenon known as ‘anaphora'. The problem of resolving pronominal and nominal anaphora to improve a system that detects drug interactions is addressed in this paper. To our knowledge, this is the first research article that tackles this issue. A corpus and a system for the evaluation of drug anaphora resolution have been developed and an analysis of the phenomena is also included. The system uses a domain-specific syntactic and semantic parser, UMLS Metamap Transfer (MMTx) [1], to select anaphoric expressions and candidate references. It is shown that a combination of the domain-specific syntax and semantic information with generic heuristics can be leveraged to produce good results comparable to other related domains. Furthermore, the analysis of the errors suggests that the use of additional semantic knowledge is needed to improve results and deal with this linguistic phenomenon in this particular domain.