Product Releases Dataset

The product releases dataset is an annotated dataset which includes ~30,000 web pages tagged by crowd-sourced labelers. Each page was labeled by 4 to 6 annotators. The number of annotators per web page can be found in the column "judgments". The annotators were asked to tag each web page with "text includes future product release" (63% pages) or "text doesn't include a future product release" (37% pages). The annotators were guided to ignore releases that happened in the past, and future releases of non products entities of mobile applications, software, movies and TV shows. Each web page is associated with labeling confidence score. The confidence score was calculated by the Appen platform and it is based on the level of agreement between multiple contributors (weighted by the contributors’ trust scores), and indicates the “confidence” in the validity of the result.