metadata
license: mit
datasets:
- kasnerz/logic2text
language:
- en
Model description
This is a fine-tuned t5-base model for generating sentences from logical forms and basic table metadata. It is a part of table-to-text pipeline with content selection and logical reasoning, trained on Logic2Text dataset. The first pipeline step is done by kategaranina/lt2t_content_selection model. For relevant code, see LT2T repository on GitHub.
Example of input:
Title: list of cities , towns and villages in vojvodina.
Header: settlement | cyrillic name other names | type | population ( 2011 ) | largest ethnic group ( 2002 ) | dominant religion ( 2002 ).
Logical form: and { only { filter_eq { all_rows ; largest ethnic group ( 2002 ) ; slovaks } } ; eq { hop { filter_eq { all_rows ; largest ethnic group ( 2002 ) ; slovaks } ; settlement } ; pivnice } }
Example of output:
pivnice is the only settlement in vojvodina with slovaks as the largest ethnic group .