Historical Negation

The foundational rule-based system that extends Negex to classify clinical findings into historical, hypothetical, and experiencer contexts. ConText uses lexical triggers like 'history of' and 'mother' combined with termination rules to define the scope of each modifier, enabling accurate patient timeline construction.

The categorical label assigned to each extracted medical entity defining its relationship to the patient. Standard classes include:

• Present: Actively affirmed in the record
• Absent: Explicitly negated
• Historical: Occurred in the past, not active
• Hypothetical: Discussed in a conditional future context
• Family History: Pertains to a relative, not the patient

The process of determining the exact span of tokens whose meaning is inverted by a negation cue. For example, in 'no evidence of acute chest pain or shortness of breath', the scope must extend across the conjunction to correctly negate both findings. Scope errors are a primary source of false negatives in extraction pipelines.

A critical disambiguation task where negation trigger words appear but do not semantically reverse a finding. Examples include:

• Double negation: 'not unlikely'
• Emphatic affirmation: 'not only pneumonia but also...'
• Rhetorical negation: 'why not treat with antibiotics' Failure to resolve pseudo-negation leads to false positive negation and lost clinical data.

A BERT-based model fine-tuned on the BioScope corpus for token-level negation and speculation detection. Unlike rule-based systems, NegBERT captures contextual embeddings that disambiguate the same word used in affirmative vs. negated contexts, achieving higher recall on complex syntactic structures.

Rigorous measurement of negation system performance requires:

• Negation Precision: Correctly flagged negations / total flagged
• Recall (Sensitivity): Correctly flagged negations / total actual negations
• False Negative Rate: Missed negations / total actual negations A high false negative rate is clinically dangerous, as it causes diseases to be incorrectly attributed to the patient's active problem list.