Clinical-sentences

Clinical-sentences performs sentence boundary disambiguation in clinical documentation using a CNN-Bi-LSTM deep learning model to address domain-specific irregularities in medical text.

Key Features:

• CNN-Bi-LSTM architecture: Combines a convolutional neural network layer with a bidirectional long short-term memory (Bi-LSTM) layer to capture local patterns and contextual dependencies in clinical text.
• Ensemble domain adaptation: Uses an ensemble approach for domain adaptation to enhance robustness across diverse clinical documentation styles.
• Training corpora: Trains on two distinct corpora to improve generalization across different formats and styles of medical records.
• Implementation framework: Implemented using the Keras neural-networks API.

Scientific Applications:

• Clinical sentence segmentation: Improves sentence segmentation and boundary disambiguation in clinical notes and other medical texts.
• Clinical NLP research: Supports research in clinical natural language processing that requires accurate sentence boundaries in domain-specific texts.

Methodology:

Implements a deep learning model combining a CNN layer with a Bi-LSTM layer and an ensemble domain-adaptation approach trained on two corpora, implemented in Keras.