Recognizing contextual polarity in phrase-level sentiment analysis
HLT '05 Proceedings of the conference on Human Language Technology and Empirical Methods in Natural Language Processing
Hot Topic Extraction Based on Timeline Analysis and Multidimensional Sentence Modeling
IEEE Transactions on Knowledge and Data Engineering
Exploiting Metrics for Similarity-Based Semantic Web Service Discovery
ICWS '09 Proceedings of the 2009 IEEE International Conference on Web Services
SentiTVchat: sensing the mood of social-TV viewers
Proceedings of the 10th European conference on Interactive tv and video
| Hi-index | 0.00 |
This paper proposes algorithms for analyzing Twitter messages. These algorithms play an important role in our 'Intelligence Circulation System', which provides several services for social TV. Twitter users often post messages about on-air TV programmes. The Intelligence Circulation System analyses these messages by using programme-related information and generates several outputs based on the analysis results. Outputs are provided to both viewers and broadcasters. The algorithms, which were designed by taking into account the characteristics of Twitter messages related to TV programmes, use auxiliary programme information, the similarity between messages, and the time series of the messages. Evaluation of our algorithms using Twitter messages about the World Cup showed that topic extraction was about 70% accurate and that message classification reached a precision of 0.85 and a recall of 0.74. Five services have been implemented in the Intelligence Circulation System by adapting the algorithms to each service.