Pharmacokinetic Table Classification (PKTC) Dataset

The annotated data stored here is for training and evaluating table classification models to identify tables containing pharmacokinetic parameter estimates and demographic information from scientific text. 

Training, validation, and test files are released in pickle format and can be read into a Pandas DataFrame. 

Each row in the DataFrame corresponds to an annotated table with the following information: 

table_id: unique identifier expressing the PubMed Open Access article and table number of the retrieved table.

pmc_link: link to the original PubMed Open Access article from which the table was retrieved.

label:  exclusive table type label (e.g. Pharmacokinetics, Demographics or Other).  

caption: the table caption, in string format. 

HTML: the raw table in HTML format (unchanged from how it appears in PubMed Open Acess XML articles). 

footer: the table footer, in string format.