Tanja Välisalo: Developing AI Tools for Studying Livestreaming / Esports

Screenshots from the first version of an analysis tool for chat data.

Livestreaming has become a significant part of game cultures through players streaming their gameplay. Especially the rapid growth of esports has made livestreaming a phenomenon that is interesting to a growing number of scholars. An essential part of the livestream experience is the chat option provided by practically every livestreaming service. The chat is simultaneously an important platform for esports audiences and their engagement with livestreams, and a window into different esports and livestreaming cultures. Understanding audience behaviours in the chat has become an important part of understanding esports livestreams in general. 

The research of chat cultures and their typical forms of communication are central to understanding how meanings and communities are formed around esports and in the digital spaces of these livestreaming platforms. Different esports and different channels have their own unique cultures, which is reflected in the chat content, the language used, and even the pace and rhythms of the chat. Different platforms also have their own distinctive features, such as custom emojis in the chat, which further diversify the content. The diversity and magnitude of data certainly provide a non-trivial challenge for research. 

Analysis tool for chat data will enable identifying unique messages as well as different forms of repetitive content in the chat.

To answer this challenge, researchers from our Centre of Excellence, namely Professor Raine Koskimaa (JYU), Professor Jaakko Peltonen (TAU) and University Teacher Tanja Välisalo (JYU) together with Project Researcher Jari Lindroos (JYU) are currently developing methods and tools for analysing esports livestreams. The goal is to create analysis tools for livestream chat data: (1) A data collection tool for Twitch chat data is in the prototype phase. A graphic interface will make it accessible for a broad group of researchers. (2) A tool for discerning formal structures of the chat and visually exploring the data is in development. (3) A tool for content analysis using machine learning methods is in development, which has thus far, focused on identifying chat content based on how unique or repetitive it is. Further development will broaden this tool to identifying and labelling content on different basis. 

Tool development is necessarily also method development. For instance, in the case of livestreaming, tool development demands and evokes discussions and decisions on what features of the chat are most relevant for researchers. Further development could also reach into connections between chat content and video stream content using technologies such as image recognition. Choices made in tool development now can also affect future directions of esports research by providing analysis tools that are broadly accessible. For these reasons, researchers in game studies and esports studies are welcomed to take part in testing and developing the tools developed here in co-operation with CoE researchers.  

Livestream chat research tools are being developed as part of FIN-CLARIAH, a joint national research infrastructure consortium in Finland. FIN-CLARIAH consists of two components, FIN-CLARIN, a previously existing infrastructure supporting research based on language data, and DARIAH-FI, launched in 2022 for developing infrastructure aimed at utilizing big data in humanities and social sciences. The current development work is funded by the Academy of Finland but is part of a Europe-wide research infrastructures DARIAH-EU and CLARIN ERIC.  

Author bio and contact info: 

Tanja Välisalo is a CoE researcher and the coordinator of the FIN-CLARIAH infrastructure project at the University of Jyväskylä.