Eventera is a real-time event recommendation system that crawls massive online media in real time, aggregates them into events, mines casual relationships among the events and generates their big map, and recommends the events to appropriate users. We build mobile/web version of Eventera, and release a real-time event recommendation system to the public in South Korea
Our contributions of Eventera are as follows:
Aggregation : Eventera crawls, aggregates and detects trending events using temporal analysis from multiple sources of media channels.
Summarization : Events are summarized as short representative sentences using centroid based summarization techniques.
Association : Sequence and interaction maps of events show casual relationship between events and affectiveness between media channels, respectively.
Screenshots of mobile version of Eventera system
Sequence map of events (left) and Interaction map of events (right)
Eventera: Real-time Event Recommendation System from Massive Heterogeneous Online Media
Dongyeop Kang, DongGyun Han, Na Hea Park, Sangtae Kim, U Kang, Soobin Lee
Submitted to IEEE International Conference on Data Mining (ICDM) 2014 (demo)
[pdf | bib]
Here is the detail list of real dataset and their links referenced in the Eventera paper. Due to privacy concerns, only sample data is available. The whole data is available with request to authors.
| Name | Volumes | Size(MB) | Description | Download |
| News | 40,786,501 | 72,704 | 26 millions of news articles are crawled from hundreds of major and minor Korean presses. The list of presses we crawl is here | sample data |
| 11,686,488 | 4,012 | 10 millions of tweets containing Korean words are crawled using Twitter API. | sample data | |
| 745,875 | 578 | 6 hundreds of thousands of Facebook posts are crawled from 23 major news Facebook pages using Facebook API. | sample data | |
| Communities | 211,349 / 216,200 / 100,359 / 259,456 |
401 / / 143 / 110 |
Postings from four major Internet Forums in South Korea (e.g., Ilbe, OU, HU) are crawled. | sample data |
| Search Queries | 3,567,185 / 3,030,078 | 1,253 / 173 | Top search queries per every ten seconds from two major search engines (e.g., Daum, Naver) in Souhth Korea are crawled using their APIs. | sample data |
| Name | Volumes | Description | Download |
| Events | 6,238 | The number of distinct events detected in Eventera database | sample data |
| SequenceMap Chains | 15,520 | Event chains whose size is larger than 2 inferred by connecting-the-dots | sample data |
