An Ensemble Semantic Textual Similarity Measure Based on Multiple Evidences for Biomedical Documents

Area of Science:

• Biomedical Informatics
• Computational Biology
• Information Science

Background:

• The rapid growth of biomedical literature necessitates efficient information retrieval.
• Accurate retrieval of data on biological entities (sequence, structure, function) is crucial for advancing biology and medicine.

Purpose of the Study:

• To develop a novel multi-evidence-based semantic text similarity calculation method for biomedical documents.
• To improve the capture of semantic information for more effective document clustering and retrieval.

Main Methods:

• Calculating MeSH-based semantic similarity.
• Calculating word embedding-based semantic similarity.
• Integrating semantic similarities using a feedforward neural network and content similarity via weighted linear combination.

Main Results:

• The proposed multi-evidence-based method demonstrated superior performance compared to existing basic methods.
• The integrated approach effectively fuses semantic and content similarities for enhanced text analysis.

Conclusions:

• The developed method significantly improves the retrieval of thematically consistent biomedical documents.
• This approach has practical applications in biological and medical research, including protein analysis and experimental design.