precision_check: Manually check positive matches
In maskedforreview/gtdnews: Online appendix for @@ paper

Description Usage Format Details Examples

100 event - article pairs that are considered as matches in the analysis presented in the paper. That is, using the preprocessing and document similarity calculation settings used in this online appendix, and using the similarity threshold of 6.54 as obtained by the threshold estimation procedure. The purpose of this data is to provide additional validation.

1	data(precision_check)

A data.frame

The other gold standard is a completely random sample of articles for which all possible matches were coded, which is required for getting a good indication of recall, but at the cost that only a limited number of positive matches (113) is obtained, and only for a limited number of articles (27). This additional validation of a random sample of 100 positive matches, for 100 unique guardian articles and 98 unique gtd events, provides a better indication of the precision. In addition, comments are provided for all matches where it's not immediately obvious that the article does or does not cover the GTD event.