The folder data contains the parallel corpora used to train the pipeline steps of Discourse Ordering, Text Structuring, Lexicalization and Referring Expression Generation.
The results obtained for all the approaches in the development and test sets of the WebNLG corpus are available on results.
In the folder human, you will be able to find all the information about the human evaluation, including the demographics of the participants (participants.json) and their ratings (grades.json).
In the folder questionaire, you will find all the information available about the qualitative analysis, including the trials (partial.txt and trials/), and the annotations (annotations.json)
The script evaluation.ipynb contains the code used to obtain all the results reported in the paper.