Data and code for the paper "B. Kleinberg, M. Mozes and I. van der Vegt, 2018. Identifying the sentiment styles of YouTube’s vloggers. To appear in Proceedings of the Conference on Empirical Methods in Natural Language Processing (EMNLP), Brussels, Belgium, 2018".
The downloaded YouTube transcripts can be found in the data/output_dir directory. We provide the transcripts come in two different ways:
- raw (
data/output_dir/raw): contains the raw and XML-encoded transcripts for each video as .txt file including temporal information about the start and end date at which a sequence is visible to the viewer when plaing a video. - parsed (
data/output_dir/parsed): contains the parsed transcripts for each video, i.e. a continuous text consisting of the concatenation of all sequences for a video.
We provide metadata information for each video in data/overview.txt. This file contains comma-separated metadata for one video in each row. The first two elements of each row uniquely define the transcript file (vlogger name + local id). For example, the transcript file for Bratayley,1 would be the file 1.txt in data/output_dir/[raw/parsed]/Bratayley. In total, the information in each row represent username of vlogger, local id, video url, view count, date of video publication, url to user's YouTube channel.