Social media platforms such as Twitter or Sina Weibo have become a common way for people to share opinions and emotions. Sports events are traditionally accompanied by strong emotions and the 2012 summer Olympic Games in London were not an exception. This project aims to develop tools for analysis of the emotions expressed in social media during the Olympic Games or other popular sports events. We use the 20 fine-grained emotion categories of GEW to allow users to distinguish the personal reactions with more details. We work both on the automatic extraction of those emotions in the tweets and on their compact visualization.
Emotion Recognition, Social Media Analysis, Emotion Visualization, Human Computation
Swiss National Science Foundation
- Renato Kempter, Valentina Sintsova, Claudiu Musat, and Pearl Pu. EmotionWatch: Visualizing Fine-Grained Emotions in Event-Related Tweets. In the 8th International AAAI Conference on Weblogs and Social Media (ICWSM), 2014
- Valentina Sintsova, Claudiu Musat, and Pearl Pu. Fine-Grained Emotion Recognition in Olympic Tweets Based on Human Computation. In Proceedings of the 4th Workshop on Computational Approaches to Subjectivity, Sentiment and Social Media Analysis (WASSA), 12–20. Atlanta, Georgia: Association for Computational Linguistics, 2013.
- Renato Kempter, Valentina Sintsova, Claudiu Musat, and Pearl Pu. Discover Emotions in Tweets & Weibos during the 2012 Olympic Games. In Video Track of the 23rd International Joint Conference on Articifial Intelligence (IJCAI). Beijing, China, 2013.
- Renato Kempter, visualization & English emotion recognition model
- Valentina Sintsova, annotation task & English emotion recognition model
- Pearl Pu, supervisor
- Claudiu Musat, data crawling & annotation task & English emotion recognition model & visualization
- Rong Hu, data crawling & Chinese emotion recognition model
- Liang Yizhong, Chinese emotion recognition model
- Marina Boia, data crawling
- Yu Chen, data crawling
We designed the visualization interface of the discovered emotions within posted tweets and weibos.
The visualization allows discovering patterns of emotion distributions evolving within the timeline of the event, along with the related messages.
Human Computation of Emotion Twitter Data
We designed a task for humans to detect the emotions of the Olympic-related tweets and provide us with emotion indicators. This task was launched on Amazon Mechanical Turk, attracting 711 workers.
The task requires workers to
- define the dominant emotion of the tweet and its strength
- detect the emotion indicators from the tweet text
- provide additional emotion indicators
The emotion model we use for annotation is based on the 20 categories of Geneva Emotion Wheel (GEW), version 2.0.
Sports-Related Emotion Corpus – SREC
The data we collected from our online task construct the Sport-Related Emotion Corpus (SREC).
After refinement, it contains 1957 tweets, with 3.48 emotion annotations in average.
Unfortunately, by Twitter terms of service, only part of the tweets is available.
You can find details on how to obtain this data for research purposes here.
Constructed Emotion Lexicon OlympLex
Our emotion recognition model is lexicon-based. Using the emotion indicators suggested in our online task, we constructed the emotion lexicon OlympLex. It contains 3193 terms each associated with distribution of 20 GEW emotion categories. It can be used to extract the emotions within the text of sport domain.
You can find details on how to obtain a copy of this lexicon for research purposes here.