Discovering event episodes from news corpora: a temporal-based approach
Proceedings of the 11th International Conference on Electronic Commerce
Early detection of buzzwords based on large-scale time-series analysis of blog entries
Proceedings of the 23rd ACM conference on Hypertext and social media
| Hi-index | 0.00 |
This paper presents a formal method of detecting emerging and changing interests that appear in document streams arriving continuously over time. Examples of such document streams include email, news articles, and weblogs (or blogs). We utilize the temporal information associated with documents in the streams and discover emerging issues and topics of interest and their change by detecting buzzwords in the documents. Buzzwords are terms that occur with strong momentum for a relatively short period of time. Our approach for buzz detection is based on the notion of "burst of activities" proposed by Kleinberg. The burst of activities is modeled using a weighted automaton. We propose an algorithm to discover buzzwords of high intensity measured by their momentum and relative duration of the bursts. The method is applied and validated on a stream of blog postings and we report the experiment results.