Applying semantic-based probabilistic context-free grammar to medical language processing--a preliminary study on...

Area of Science:

• Natural Language Processing
• Computational Linguistics
• Medical Informatics

Background:

• Semantic-based sublanguage grammars are effective for medical language processing.
• Grammar-based parsers often face sentence ambiguity, leading to multiple parse trees.
• Augmenting grammars with probabilities is an underexplored solution for ambiguity resolution.

Purpose of the Study:

• To investigate the use of probabilistic context-free grammars (PCFG) for reducing parsing ambiguity in medical language.
• To evaluate the performance of a semantic-based PCFG for medication finding extraction.

Main Methods:

• Developed a semantic-based Context-Free Grammar (CFG) for parsing medication sentences.
• Created a Treebank of 4564 medication sentences from discharge summaries.
• Derived a semantic-based PCFG from the Treebank and evaluated it using 10-fold cross-validation.

Main Results:

• The PCFG parser demonstrated a significant improvement in parsing performance compared to the CFG parser.
• Probabilistic augmentation effectively reduced ambiguity in parsing medication-related sentences.
• The developed PCFG achieved higher accuracy in medication extraction tasks.

Conclusions:

• Probabilistic context-free grammars are a viable and effective method for resolving ambiguity in medical natural language processing.
• This approach enhances the accuracy of medication extraction from clinical text.
• Further research into probabilistic grammars can advance medical language understanding.