Burst: real-time events burst detection in social text stream

Area of Science:

• Computer Science
• Data Science
• Social Media Analytics

Background:

• The proliferation of social media generates vast, noisy text streams, making early event detection challenging.
• Extracting meaningful information and identifying emerging phenomena from this data requires effective filtering and representation techniques.
• Twitter data, in particular, presents challenges due to its sparse and noisy nature, hindering accurate burst event detection.

Purpose of the Study:

• To present a concise and effective approach for classifying and detecting bursty events within social media text streams.
• To explore methods for cleaning and representing noisy text data for improved event detection.
• To identify and track event keywords and patterns based on relevant features for real-time analysis.

Main Methods:

• Development of a novel approach for bursty event detection in social text streams.
• Implementation of text cleaning and profound representation techniques for noisy data.
• Classification and detection of event keywords using a feature-based model to identify booming patterns.

Main Results:

• The proposed approach demonstrates proficiency in detecting valuable patterns of interest within social media data.
• Experimental results show superior performance compared to state-of-the-art methods in extracting burst events.
• The feature-based model successfully identifies and tracks event patterns across different time spans.

Conclusions:

• The developed approach offers an effective solution for bursty event detection in noisy social media text streams.
• Accurate identification of event keywords and patterns is crucial for understanding real-time phenomena.
• This research contributes to advancing the field of social media analytics and event detection.