A Domain Knowledge-Enhanced LSTM-CRF Model for Disease Named Entity Recognition

Area of Science:

• Biomedical Natural Language Processing (NLP)
• Bioinformatics
• Computational Biology

Background:

• Disease Named Entity Recognition (NER) is crucial for biomedical NLP applications.
• Accurate disease extraction aids patient profiling, clinical trial matching, and information retrieval for clinicians.
• Existing methods require enhancement for precise disease identification in scientific literature.

Purpose of the Study:

• To develop and evaluate a novel domain knowledge-enhanced deep learning model for disease NER.
• To improve the accuracy of identifying disease entities in biomedical texts.
• To enhance the performance of information retrieval and clinical applications through better disease annotation.

Main Methods:

• A Long Short-Term Memory network-Conditional Random Field (LSTM-CRF) model enhanced with domain knowledge was proposed.
• Character-level Convolutional Neural Network (CNN) and character-level LSTM were integrated for input embedding.
• The model was trained and evaluated on a scientific article dataset.

Main Results:

• The proposed domain knowledge-enhanced LSTM-CRF model demonstrated superior performance in disease NER.
• Experimental results showed significant improvements compared to existing state-of-the-art methods.
• The integration of character-level CNN and LSTM embeddings contributed to enhanced recognition accuracy.

Conclusions:

• The developed model effectively addresses the challenges in disease NER within biomedical literature.
• Domain knowledge integration and advanced deep learning architectures are key to improving NER performance.
• This work offers a valuable tool for advancing biomedical NLP and related applications.