Negation and speculation processing: A study on cue-scope labelling and assertion classification in Spanish clinical...

Area of Science:

• Computational linguistics
• Artificial intelligence in healthcare
• Biomedical informatics

Background:

• Healthcare data analysis relies on Natural Language Processing (NLP) for pattern discovery in vast text volumes.
• Accurate processing of negation and speculation is crucial for reliable NLP in clinical settings.

Purpose of the Study:

• To address challenges in processing negation and speculation within clinical text using deep learning.
• To propose a method for converting cue-scope annotations to assertion annotations, overcoming data limitations.
• To evaluate deep learning models, particularly Transformer-based approaches, on clinical Spanish text.

Main Methods:

• Utilized state-of-the-art deep learning, focusing on cue-scope labeling and assertion classification.
• Developed a methodology for automatic conversion of cue-scope to assertion annotations.
• Conducted experiments with varying training data and adversarial test examples.

Main Results:

• Transformer-based models demonstrated a clear advantage over baselines and prior work.
• The proposed annotation conversion methodology proved effective.
• Models achieved superior performance on the NUBes clinical Spanish text corpus.

Conclusions:

• Deep learning, especially Transformer models, significantly advances NLP for clinical text analysis.
• The developed methods improve the handling of negation and speculation, crucial for healthcare applications.
• This work offers a pathway to more robust and accurate interpretation of clinical narratives.