A Gated Dilated Convolution with Attention Model for Clinical Cloze-Style Reading Comprehension

Area of Science:

• Medical Informatics
• Natural Language Processing
• Artificial Intelligence

Background:

• Machine comprehension in clinical medicine is valuable but underdeveloped.
• Existing cloze-style reading comprehension models are often time-consuming.
• There is a need for efficient and effective clinical machine reading comprehension models.

Purpose of the Study:

• To address the limitations of existing models in clinical machine reading comprehension.
• To propose an efficient and effective model for cloze-style machine reading comprehension in the medical domain.
• To improve the performance and speed of clinical machine reading comprehension.

Main Methods:

• Development of a novel Gated Dilated Convolution with Attention (GDCA) model.
• The GDCA model incorporates a gated dilated convolution module and an attention mechanism.
• The model is designed for high parallelism and capturing long-distance dependencies.

Main Results:

• The GDCA model achieved state-of-the-art results on the CliCR dataset.
• Our model surpassed existing best models across several performance metrics.
• The training speed of the GDCA model was 8 times faster than the previous best model.

Conclusions:

• The proposed GDCA model offers a significant advancement in clinical machine reading comprehension.
• The model demonstrates superior performance and remarkable efficiency.
• GDCA holds great potential for practical applications in clinical medicine.