Fast and effective biomedical named entity recognition using temporal convolutional network with conditional random...

Area of Science:

• Computational biology
• Bioinformatics
• Natural Language Processing

Background:

• Biomedical named entity recognition (Bio-NER) is crucial for extracting knowledge from biomedical literature.
• Current state-of-the-art Bio-NER models, such as BiLSTM and BERT, are computationally intensive.
• There is a need for more efficient Bio-NER models.

Purpose of the Study:

• To propose a computationally efficient Temporal Convolutional Network with a Conditional Random Field (TCN-CRF) layer for Bio-NER.
• To enhance the TCN model by fusing features from convolutional kernels of different sizes.
• To evaluate the performance of the proposed TCN-CRF model against existing deep learning models.

Main Methods:

• Utilized a Temporal Convolutional Network (TCN) for feature extraction in Bio-NER.
• Integrated a Conditional Random Field (CRF) layer for decoding and obtaining the final Bio-NER results.
• Improved the TCN by incorporating fused features from convolutional kernels with varying sizes.

Main Results:

• The proposed TCN-CRF model demonstrated comparative performance to five other deep learning models on the GENIA and CoNLL-2003 datasets.
• The TCN-CRF model achieved this performance with significantly less training time compared to existing methods.
• The implemented code is publicly available for the research community.

Conclusions:

• The TCN-CRF model offers an efficient and effective solution for biomedical named entity recognition.
• The proposed feature fusion strategy enhances Bio-NER performance within the TCN framework.
• This work contributes a computationally lighter yet powerful alternative for Bio-NER tasks.