Logic2Text subset

Logic2Text features open–domain Wikipedia tables manually annotated with descriptions of common logic types and their underlying logical forms for the T2T task.

Images of tables are stored in the .zip folders, while textual formats of tables and all metadata are availabe in .json files.

Data Fields

instance_id - (str) unique ID of a given instance;
topic - (str);
wiki - (str) URL of the Wikipedia page;
url - (str);
action - (str);
sent - (str) gold statement;
annotation - (dict);
logic - (dict);
logic_str - (str) logical form;
interpret - (str);
num_func - (str);
nid -(str);
g_ids - (dict);
g_ids_features - (dict);
g_adj - (dict);
table_header - (List[str]) table header names;
table_cont - List[List[str]])table row values;
table_id - (str) unique ID of a given table;
title - (str) table title;
html_content - (str) source HTML page of a table;
snapshot_timestamp - (str) timestamp of the HTML page version in Wikipedia history;
filename - (str);
image_name - (str) table image name;
table_xml - (str) table in XML format;
table_html - (str) table in HTML format;
table_latex - (str) table in LaTeX format.