Effective Event Identification in Social Media
0 likes339 views
This document summarizes a method for effective event identification in social media. It describes using clustering to group similar social media documents into events, both for known and unknown events. Key aspects of the method include using multiple document features like text, hashtags and timestamps. An ensemble algorithm learns weights for these features. Techniques like considering bursty vocabulary terms and time decay help improve clustering effectiveness over time. An evaluation on real-world event data from Flickr found the method using bursty vocabulary and time decay obtained the highest quality results.
Effective Event Identification in Social Media
- 1. + Effective Event Identification in Social Media 2014/4/27(Mon.) Chang Wei-Yuan @ MakeLab Lab Meeting Fotis Psallidas, Hila Becker DEB’13
- 2. + Outline n Introduction n Method n Known-Event Identification n Unknown-Event Identification n Improving Identification Effectiveness n Experimental Evaluation n Conclusion n Thought 2
- 3. + Introduction nOnline social media are extensively distribute content related to real-world events. n event: something that occurs at a certain time in a certain place 3
- 4. + Introduction nOnline social media are extensively distribute content related to real-world events. n event: something that occurs at a certain time in a certain place 4 Goal：Identifying Events and Associated Social Media Documents
- 5. + Introduction nGeneral approach: group similar documents via clustering n Each cluster corresponds to one event and its associated social media documents 5
- 6. + Introduction nChallenges n Uneven data quality n Highly heterogeneous n Dynamic data stream of event information n Number of events unknown 6
- 7. + Event Identification nKnown-Event Identification nUnknown-Event Identification nImproving Identification Effectiveness 7
- 9. + Known-Event Identification nSocial media content related to known events n reside in multiple social media sites, each contributing different information nTo retrieve cross-site social media documents for same event n miss many relevant event documents 9
- 10. + Known-Event Identification nIn the first step, using the known event properties to achieve high-precision results. nIn the second step, using term extraction and frequency analysis to improve recall. 10
- 11. + Unknown-Event Identification 11 na Twitter stream may contain many tweets related to an event n with messages related to other events n with messages unrelated to events
- 12. + Unknown-Event Identification nThe proposed online clustering framework n leverages the multiple features to decide when two social media documents correspond to the same event 12
- 13. Social Media Document Clustering Framework Document feature representation Social media documents Event clusters 13
- 14. Ensemble Algorithm nThe proposed online clustering framework n deployed ensemble learning methods to learn and associate each feature with a weight and a threshold that capture the importance of the features 14
- 15. Consensus Function: combine ensemble similarities Wtitle Wtags Wtime 15 f(C,W) Ctitle Ctag s Ctime Ensemble clustering solution Learned in a training step Ensemble Algorithm
- 16. Event Classification nThe proposed online clustering framework n deployed event classification to distinguish between event-related clusters and non-event ones 16
- 17. Event Classification 17 Ensemble clustering solution Event unrelated to events related to an event event classification
- 18. + Improving Identification Effectiveness nHow events behave over time have a significant impact on the effectiveness of the document clustering procedure? nHow to refine the clustering procedure to benefit from these factors is a challenging task? 18
- 19. + URLs nURLs in event-related social streams are ubiquitous. Individuals use them to share meaningful event-related external content. 19
- 20. + Bursty Vocabulary nThe social media content related to an event tends to revolve around a central topic. n this central topic is expressed by a set of terms that is significantly more frequent n span a wide time range exhibit a different set of these bursty terms at different points of their lifetime. 20
- 21. + Bursty Vocabulary nThe social media content related to an event tends to revolve around a central topic. 21
- 22. + Time Decay na time decay function to the clustering framework n penalizes clusters that have been inactive for a long time. n re-triggers events that have been inactive for some time if the similarity score without the time-decay factor is strong enough. 22
- 23. + Experimental Evaluation nData n Upcoming dataset n 273,842 multi-featured Flickr photos that correspond to 9,613 real-world events from the Upcoming event. nthe BurstyV + TimeDec technique obtained the highest quality results. 23
- 24. + Conclusion nThis article discussed the event identification task under two different scenarios, known- and. nWe showed how to identify event content effectively n how we can exploit rich features of the social media documents n revealing temporal patterns of the relevant content 24
- 25. + Thought 25
- 26. + Thanks for listening. 2014 / 4 / 27(Mon.) @ MakeLab Group Meeting v123582@gmail.com
Editor's Notes
- #2: Fotis Psallidas: Columbia University Hila Becker: Google, Inc.
- #9: For instance, YouTube might contain videos for the Super Bowl event, whereas Twitter users might discuss the event by sharing short text messages, or tweets. Such highly specific queries tend to retrieve event-related documents with high precision but with low recall.
- #15: We can cluster out document collection according to the variety of feature reps. discussed, each would have its own