MaTED: Metadata-Assisted Twitter Event Detection System

Abstract

Due to its asynchronous message-sharing and real-time capabilities, Twitter offers a valuable opportunity to detect events in a timely manner. Existing approaches for event detection have mainly focused on building a temporal profile of named entities and detecting unusually large bursts in their usage to signify an event. We extend this line of research by incorporating external knowledge bases such as DBPedia, WordNet; and exploiting specific features of Twitter for efficient event detection. We show that our system utilizing temporal, social, and Twitter-specific features yields improvement in the precision, recall, and DERate on the benchmarked Events2012 corpus compared to the stateof-the-art approaches.

Social Media Event Detection Using Spacy Named Entity Recognition and Spectral Embeddings

“…We used spaCy's pretrained NER model [15], a transition-based parser, to extract key phrases. Each named entity was assigned a "burstiness" score within each time window corresponding to how often it appeared in that window relative to every other window [1]. This score is multiplied by the log of the number of followers, favorites, and retweets to weight tweets by engagement similar to MaTED's approach.…”

“…Twitter is an important social media platform for this purpose, given its ongoing popularity, volume of publicly accessible data, and concise format. As natural language processing (NLP) methods become more sophisticated, discovering more robust, nuanced, and scalable approaches to general event detection via Twitter continues to be an active field of research [1]- [6].…”

“…Here, we build off an existing approach, Metadata-Assisted Twitter Event Detection (MaTED) [1]. The MaTED approach consists of four main components, including detection of important phrases from tweets, creating temporal profiles of these phrases to identify "bursty" phrases, clustering bursty phrases with an aim to group related phrases about an event, and characterizing an event from the clusters obtained.…”

“…Fig.1: Twitter event detection pipeline. Adapted from MaTED[1], except spaCy is used for named entity recognition, a transformerbased similarity measure is used during graph formation, and the graph is embedded in a lower-dimensional space before clustering.…”

Make Tweets Great Again: Who Are Opinion Leaders, and What Did They Tweet About Donald Trump?

“…In 2005, only 5% of American adults used social media; currently, about seven of 10 Americans use at least one social media site (Pew Research Center, 2019). With 186 million daily users (Twitter Inc., 2020) producing 500 million messages per day and over 15 billion Application Programming Interface (API) calls per day (Pandya et al, 2020), Twitter remains a key part of the social media landscape and the media diet of many Americans. Twitter gained new prominence when the former President of the United States, Donald J. Trump, adopted Twitter as his preferred platform to make official statements directly to the public (Landers, 2017).…”