LEDS: Local Event Discovery and Summarization from Tweets

Open Webpage

Zhi Liu, Yan Huang, Joshua R. Trampier

Published: 01 Nov 2016, Last Modified: 15 May 2025Proceedings of the 24th ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems (SIGSPATIAL ’16)EveryoneCC BY-ND 4.0
Abstract: We present LEDS, a framework for discovering and summarizing local events in Twitter streams. LEDS integrates spatiotemporal clustering with keyword extraction to detect event candidates, then applies summarization (centroid-based ranking + language-model scoring) to generate concise descriptions. Evaluation on real tweets shows superior recall and summary quality compared to baselines.