Toward information extraction: identifying protein names from biological papers

Area of Science:

• Biotechnology
• Bioinformatics
• Computational Biology

Background:

• Understanding gene expression and product interactions is crucial for deciphering life phenomena.
• Vast amounts of biological knowledge are locked in published articles, necessitating advanced information extraction (IE) systems.
• Existing IE methods struggle with novel gene and protein names (coinages) common in biomedical literature.

Purpose of the Study:

• To develop a robust method for identifying material names, specifically genes and proteins, within biomedical texts.
• To overcome the limitations of dictionary-based approaches in recognizing newly coined terms.
• To enhance the accuracy of information extraction from scientific literature.

Main Methods:

• Propose a new information extraction method named PROPER.
• Utilize surface clues within character strings to identify potential material names.
• Develop a system capable of detecting both previously known and newly defined terms.

Main Results:

• The PROPER method achieves high precision of 94.70% in extracting material names.
• The PROPER method demonstrates a high recall rate of 98.84%.
• The system effectively identifies material names irrespective of their prior definition or common usage.

Conclusions:

• PROPER offers a significant advancement in extracting gene and protein names from scientific literature.
• This method enhances the capability of intelligent information extraction systems for biological research.
• The approach successfully addresses the challenge of identifying novel terms in biomedical documents.