Long COVID SDOH NER

This dataset contains text for over 7,000 case report sections in academic papers related to Post COVID-19 Condition (PCC), also known as Long COVID. It is comprised of three columns: 'Token,' 'entity_type,' and 'case_report_id.' There are over 3.4 million observations in total.

Entity types are presented in CoNLL format and relate to sociodemographic, behavioral and selected biomedical dimensions relevant to PCC. There are 27 core entity types (53 in CoNLL): ‘Access_To_Care,’ ‘Age,’ ‘Condition,’ ‘Diet,’ ‘Disability,’ ‘Education,’ ‘Employment,’ ‘Exercise,’ ‘Family_Member,’ ‘Gender,’ ‘Geographic_Entity,’ ‘Housing,’ ‘Income,’ ‘Insurance_Status,’ ‘Language,’ ‘Marital_Status,’ ‘Mental_Health,’ ‘Race_Ethnicity,’ ‘Severity,’ ‘Sexual_Orientation,’ ‘Social_Support,’ ‘Spiritual_Beliefs,’ ‘Substance,’ ‘Treatment,’ ‘Vaccine,’ ‘Violence_Or_Abuse,’ and ‘O’ (non-entity).

The entities were annotated with a fine-tuned BERT-base-uncased model, which was trained on a subset of 501 case reports and 3000 sets of synthetic sentences with diverse variations of each entity type.

For reproducibility, an additional file with case report sections alone is included.