• Description:

Wikipedia tables with at least 3 rows and 2 columns, 3 random rows for each table were selected for further annotation. Each row was annotated by a different person, so the dataset is composed by (one row table, text description) pairs. Annotations include at least 2 cells of the row, but do not require to include them all. The dataset follows a standarized table format.

Split Examples
'test' 2,000
'train' 10,000
'validation' 1,318
  • Feature structure:
    'input_text': FeaturesDict({
        'table': Sequence({
            'column_header': string,
            'content': string,
            'row_number': int16,
    'target_text': string,
  • Feature documentation:
Feature Class Shape Dtype Description
input_text FeaturesDict
input_text/table Sequence
input_text/table/column_header Tensor string
input_text/table/content Tensor string
input_text/table/row_number Tensor int16
target_text Tensor string
  • Citation:
  title={Table-to-Text: Describing Table Region with Natural Language},
  author={Junwei Bao and Duyu Tang and Nan Duan and Zhao Yan and Yuanhua Lv and Ming Zhou and Tiejun Zhao},