SBLC: a hybrid model for disease named entity recognition based on semantic bidirectional LSTMs and conditional...

Area of Science:

• Biomedical Informatics
• Natural Language Processing
• Machine Learning

Background:

• Disease Named Entity Recognition (NER) is crucial for processing medical texts.
• Accurate disease NER is challenging due to complex modifiers and remains an open research problem.
• Effective NER is essential for medical information extraction and text mining.

Purpose of the Study:

• To propose a novel hybrid model for disease Named Entity Recognition (NER).
• To evaluate the model's performance against state-of-the-art methods.
• To develop a disease NER model that is easily applicable to medical text processing.

Main Methods:

• A hybrid model, Semantics Bidirectional LSTM and CRF (SBLC), was developed.
• The SBLC model integrates word embeddings, Bidirectional Long Short-Term Memory networks, and Conditional Random Fields.
• The model was evaluated on the NCBI disease dataset and compared against nine baseline methods.

Main Results:

• The SBLC model achieved a high F1 score of 0.862.
• The proposed SBLC model outperformed all nine state-of-the-art baseline methods.
• The model demonstrated effectiveness in identifying disease entities in medical texts.

Conclusions:

• The SBLC model demonstrates superior performance in disease NER.
• The model's independence from external domain dictionaries enhances its practical applicability.
• This research contributes an effective solution for disease NER in medical text processing.