This dataset was created during my PhD (http://www.tdg-seville.info/fogallego/Personal%20Info) at the University of Seville. We didn't found any datasets with labelled conditions so we decided to build one since our main goal for the PhD was to be able to identify conditions without relying on user-defined patterns or requiring any specific-purpose dictionaries, taxonomies, or heuristics.

We presented this dataset in a poster session during Machine Learning Summer School Madrid 2018 (http://mlss.ii.uam.es/mlss2018/posters.html).

Content

The reviews in English and Spanish were randomly gathered from ciao.com between April 2017 and May 2017. The sentences were classified into 15 domains according to their sources, namely: adults, baby care, beauty, books, cameras, computers, films,
headsets, hotels, music, ovens, pets, phones, TV sets, and video games.

Our dataset consist of two files: sentences.csv and conditions.csv. The first one contains the whole set of sentences and the second one the manually labelled conditions.

In order to better understand the meaning of each column, I'll explain them in detail:

sentence.csv:

• sentence_uuid: the unique identifier of the sentence

• sentence_text: the text of the sentence

• language: the language of the sentence

• domain: the domain of the sentence

• labelled: whether the sentence was labelled or not

conditions.csv:

• sentence_uuid: the unique identifier of the corresponding labelled sentence

• condition_uuid: the unique identifier of the condition

• begin_connective: the character position where the connective of the condition starts

• end_connective: the character position where the connective of the condition ends

• begin_condition: the character position where the rest of the condition starts

• end_condition: the character position where the rest of the condition ends

• language: the language of the corresponding labelled sentence

• domain: the domain of the corresponding labelled sentence

Acknowledgements

My PhD and this dataset were supported by Opileak.com and the Spanish R&D programme (grants TIN2013-
40848-R and TIN2013-40848-R).