Table 3 Comparison of predicted labels by best performing baselines: BioBERT, BioBERT + CRF and our NLP model for running examples. The comparison highlights the challenges associated with entity identification, including correctly identifying an entity, failing to identify an entity type, and misclassifying a non-entity
Sentence with ground truth label | Entity Category | Our Approach | BioBERT | BioBERT + CRF |
|---|---|---|---|---|
The patient has diabetes and is a smoker | Disease, Smoking | diabetes, smoker | diabetes, smoker | diabetes, smoker |
There were more unemployed men than women | Employment, Gender | unemployed, men, women | men, women | men |
The individual's relationship status is single and currently employed | Relationship Status, Employment | single, employed | - | - |
The patient has hypertension and engages in regular physical activity | Disease, Lifestyle Factors | hypertension, physical activity | hypertension | hypertension |
The individual has completed a higher level of education and has a specific disease | Socioeconomic Factors, Disease | education | disease | disease |
The event occurred on a specific date and involved healthcare interaction | Temporal Factors, Healthcare System Interaction | date, healthcare interaction | date | date |
The patient has a psychological condition | Psychological Condition | psychological condition | psychological condition | psychological condition |
The patient’s diet affects their blood pressure | Lifestyle Factors | diet | blood | blood |
The individual’s income level is associated with a specific disease | Socioeconomic Factors, Disease | income level, disease | disease | disease |