Improving the CONTES method for normalizing biomedical text entities with concepts from an ontology with (almost) no...

Area of Science:

• Information Extraction
• Natural Language Processing
• Ontology Engineering

Background:

• Entity normalization links text expressions to ontology concepts.
• Machine learning methods require large datasets, limiting specialized domain applications.
• Existing CONTES method has limitations in scalability and handling unseen terms.

Purpose of the Study:

• To improve entity normalization in specialized domains.
• To address limitations of the CONTES method, including scalability and out-of-vocabulary terms.
• To enhance the accuracy and applicability of concept linking.

Main Methods:

• Assessing dimensionality reduction techniques for ontology representations.
• Calibrating model parameters for improved prediction accuracy.
• Developing a specific method to handle out-of-vocabulary words.

Main Results:

• Proposed methods aim to reduce dimensionality and improve prediction accuracy.
• A strategy is introduced to manage out-of-vocabulary words effectively.
• The study seeks to overcome CONTES's scalability and generalization issues.

Conclusions:

• The proposed enhancements are expected to improve entity normalization performance in specialized domains.
• Addressing dimensionality, parameter calibration, and OOV words will lead to more robust concept linking.
• This work contributes to more effective information extraction from specialized texts.