Context-aware multi-token concept recognition of biological entities.

Area of Science:

• Bioinformatics
• Computational Biology
• Natural Language Processing

Background:

• Concept recognition, encompassing named entity recognition and normalization, is crucial in bioinformatics.
• Traditional dictionary-based methods struggle with concept variation, especially for multi-token entities.
• Existing methods show degraded performance on recognizing complex biological terms in literature.

Purpose of the Study:

• To develop an improved concept recognition method for multi-token biological entities.
• To address the limitations of conventional approaches in handling concept variation.
• To enhance the accuracy of identifying biological entities in scientific texts.

Main Methods:

• Utilizing neural models combined with literature contexts for concept recognition.
• Leveraging contextual information from biological knowledge bases for concept normalization.
• Integrating concept normalization prior to the named entity recognition process.

Main Results:

• The proposed model demonstrates improved performance compared to conventional methods.
• Significant performance gains were observed particularly for multi-token concepts with high variation.
• The method effectively handles the variability of biological concepts in literature.

Conclusions:

• The developed model offers effective concept recognition for bioinformatics.
• The approach is expected to benefit various natural language processing tasks in the field.
• This method advances the accurate processing of biological information from scientific literature.