BioInfer: a corpus for information extraction in the biomedical domain

Area of Science:

• Biomedical Informatics
• Computational Biology
• Natural Language Processing

Background:

• Growing interest in applying information extraction to the biomedical domain.
• Need for annotated corpora to develop and evaluate gene, protein, and RNA relationship extraction methods.

Purpose of the Study:

• Introduce BioInfer, a novel annotated corpus for biomedical English.
• Provide a resource for developing and evaluating information extraction systems.

Main Methods:

• Developed an annotation scheme for named entities and their relationships.
• Included dependency analysis of sentence syntax.
• Created ontologies for entity and relationship types.

Main Results:

• Corpus contains 1100 sentences from biomedical abstracts.
• Annotations include relationships, named entities, and syntactic dependencies.
• Unique in combining annotation types and detail level for relationship annotation.

Conclusions:

• BioInfer corpus targets protein, gene, and RNA relationships.
• Serves as a resource for developing information extraction systems, parsers, and domain analyzers.
• Corpus will be maintained and further developed.