Improving the dictionary lookup approach for disease normalization using enhanced dictionary and query expansion.

Area of Science:

• Biomedical informatics
• Natural Language Processing
• Information Retrieval

Background:

• The exponential growth of biomedical literature necessitates automated methods for disease mention recognition and normalization.
• Existing approaches often rely on Conditional Random Fields (CRFs) for recognition and dictionary lookup for normalization.
• Improving the precision and effectiveness of disease-based information retrieval is crucial.

Purpose of the Study:

• To develop an automated system for recognizing and normalizing disease mentions in biomedical text.
• To investigate techniques for enhancing dictionary lookup-based disease normalization.
• To evaluate the performance of the developed CRF-based model against existing methods.

Main Methods:

• Developed a Conditional Random Fields (CRF)-based model for automated disease mention recognition.
• Applied and evaluated various techniques to improve dictionary lookup-based disease normalization.
• Utilized the BioCreative V CDR track dataset for performance evaluation and comparison.

Main Results:

• The developed CRF-based model achieved an F-measure of 0.77 for disease normalization.
• This performance significantly outperformed the best dictionary lookup-based baseline method by an F-measure of 0.13.
• The study demonstrated the effectiveness of the integrated CRF and enhanced dictionary lookup approach.

Conclusions:

• The proposed CRF-based model offers a robust solution for automated disease recognition and normalization.
• Enhanced dictionary lookup techniques substantially improve normalization accuracy in biomedical text.
• This work contributes to more precise and effective disease-based information retrieval from large-scale literature.