Interactive Learning for Identifying Relevant Tweets to Support Real-time Situational Awareness.

Area of Science:

• Computer Science
• Information Science
• Social Media Analytics

Background:

• Real-time social media data is crucial for situational awareness.
• High data noise and user-specific relevancy definitions complicate information extraction.
• Existing methods lack real-time, user-interactive retraining for dynamic needs.

Purpose of the Study:

• To develop an interactive learning framework for real-time short text relevance classification.
• To enable on-the-fly model retraining based on user feedback for immediate improvements.
• To enhance situational awareness by accurately classifying dynamic social media data.

Main Methods:

• Developed a novel interactive learning framework for iterative user correction.
• Integrated the framework with the Social Media Analytics and Reporting Toolkit (SMART) 2.0.
• Computationally evaluated the model's ability to learn at interactive rates.

Main Results:

• The proposed framework significantly outperforms state-of-the-art machine learning models.
• Demonstrated effective real-time classification and immediate predictive improvements.
• Successful integration with SMART 2.0 for visual analytics in situational awareness.

Conclusions:

• The interactive learning framework enhances real-time social media data analysis for situational awareness.
• User-driven, on-the-fly model adaptation is effective for dynamic information needs.
• The framework shows promise for applications requiring rapid and accurate classification of streaming data.