Piecing together the puzzle: Improving event content coverage for real-time sub-event detection using adaptive...

Area of Science:

• Social Media Analysis
• Information Science
• Computer Science

Background:

• Social media provides timely event information, but traditional monitoring systems struggle with comprehensive sub-event detection.
• Existing methods often miss crucial event details due to incomplete data collection, hindering real-time analysis.

Purpose of the Study:

• To propose a novel framework, Sub-event detection by real-TIme Microblog monitoring (STRIM), for improved real-time event sub-detection and summarization.
• To enhance the accuracy and completeness of event monitoring by addressing limitations in current data collection and analysis.

Main Methods:

• Implemented an adaptive microblog crawler to increase event data coverage while minimizing irrelevant content.
• Developed a real-time stream division methodology for temporal feature analysis using a burst detection algorithm.
• Extracted and recombined content features from divided streams for final sub-event summarization.

Main Results:

• The STRIM framework demonstrated significant improvements in event recall (44.44%) and event precision (9.57%) compared to traditional methods.
• Adaptive crawling and stream division/recombination techniques effectively identified additional valid sub-events.
• Improved data quality and coverage directly contribute to more accurate and comprehensive event detection.

Conclusions:

• The proposed STRIM framework offers a more effective approach to monitoring and understanding real-world events through social media.
• By capturing a more complete set of sub-events, the framework provides a deeper understanding of complex events.
• STRIM advances the field of event detection by enhancing real-time analysis and data comprehensiveness.