Published 21 October 2020

Exploring Livestream Chat

Researchers from Digital Creativity Labs, the Department of Computer Science, and IGGI have presented a large-scale dataset of video game livestream chat, consisting of over 60 million tokens.

The paper was presented at the AAAI Conference on Artificial Intelligence and Interactive Digital Entertainment (AIIDE 2020) which this year took place in a virtual format.

Livestream chat is becoming increasingly popular and provides valuable insights into the way in which the audience is engaging with the stream. However this is not a straightforward task because livestream chat is a rich and complex domain.

Tokens (words, symbols, abbreviations, emotes,  emoji and emoticons) commonly used in livestream chat are often unknown to traditional language understanding efforts that focus on straightforward text.

As part of the investigation, researchers presented a case study analysis of word vector methods applied to the dataset. The work demonstrated that whilst the vector space is strangely shaped, it is also clusterable and that the resulting clusters correlate with features such as streamer popularity.

The research team hopes that the livestream chat dataset, the discussion of its unique features, and the challenges highlighted for future work will invigorate the research community into further study of livestream chat.