Trans-related Online Corpus (TROC)

The dataset is classified according to the stance taken by commenters in YouTube and Reddit towards transgender people, categorised by 7 human annotators as pro-trans, contra-trans, two-sided/neutral, and unclear. To assess the level of reliability of annotation, the same comments were labeled by two annotators independently. The entire dataset was annotated by 7 annotators divided into 4 different pairs. Prior to the annotation task, all annotators were given annotation guidelines that briefly summarised the criteria of annotation. Overall, the agreement among annotators could be classified as "moderate." The dataset includes three corpora. 0. Raw corpus of YouTube and Reddit comments (see readme.txt). 1. Stance annotated corpus of contra-trans and pro-trans comments (contra-trans comments - 1264 pro-trans comments - 1023). 2. Stance annotated corpus part of speech tagged. FILES No Number of comments Words 01_youtube 659 14298 02_youtube 568 13704 Total: 1227 28002 No Number of comments Words 01_reddit 240 14152 02_reddit 1167 50498 03_reddit 58 1540 07_reddit 877 40487 Total: 2342 106677 Reference: J. Ruzaitė et al. (2026) "Polarisation and Stance-Taking across Platforms: Challenges of Stance Annotation in Trans-Related Debates"