Assessment of approximate string matching in a biomedical text retrieval problem.

Area of Science:

• Biomedical Informatics
• Computational Biology
• Text Mining

Background:

• Text-based search is crucial for biomedical data mining and knowledge discovery.
• Character errors in literature can significantly impact the accuracy of data mining processes.
• Developing robust methods to address these errors is essential for reliable biomedical research.

Purpose of the Study:

• To evaluate the effectiveness of the Smith-Waterman algorithm with affine gap penalty for biomedical literature retrieval.
• To assess the algorithm's performance in matching medicinal herb names between herbal medicine and medicinal chemistry literature.
• To determine the optimal string identity level for accurate name matching.

Main Methods:

• The Smith-Waterman algorithm with affine gap penalty was employed for sequence alignment.
• Names of medicinal herbs were extracted from herbal medicine and medicinal chemistry literature.
• The algorithm was tested at various string identity levels ranging from 80% to 100%.

Main Results:

• The Smith-Waterman algorithm demonstrated utility in biomedical text retrieval.
• Optimal performance was achieved at an 88% string identity level.
• At 88% string identity, the algorithm yielded a recall of 96.9% and a precision of 97.3%.

Conclusions:

• The Smith-Waterman algorithm is a valuable tool for enhancing the accuracy of biomedical text retrieval.
• This approach can effectively improve the success rate of data mining in biomedical literature.
• The findings support the use of sequence alignment algorithms for overcoming character errors in scientific texts.