Cross-Media Event Coreferencing for Fast, Reliable Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems struggle to accurately and efficiently extract event attributes from diverse media types like social media and news articles, due to differences in text length, reporting style, and reliability, which hinders timely decision-making.
Innovation Solution
A system and method for cross-media event detection and coreferencing that includes modules for filtering, organization, clustering, and verification of social media data, using modules like ingestion, filtering, organization, clustering, and verification to identify and verify events across multiple media types.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If event information is extracted from social media, then breaking news can be detected quickly, but the information is often unreliable and limited
Solution Approach 1:
The system merges event extraction from multiple media types (social media, news articles, blogs, forums) into a unified event detection framework. By combining fast social media data with reliable news articles and other sources, the system achieves both rapid breaking news detection and high information reliability through multi-source verification
Solution Approach 2:
The system implements feedback mechanisms where extracted event information is continuously verified against multiple sources. Event attributes are cross-checked across different media types, and verification results feed back into the extraction process to improve accuracy and reliability of detected events
2Reliability
If event information is extracted from news articles, then reliable and rich context is provided, but breaking news is reported slower
Solution Approach 1:
The system performs preliminary action by continuously monitoring and pre-processing news articles and other media sources before events need to be detected. Event extraction pipelines are pre-configured and ready to process incoming data streams, enabling rapid response when breaking news occurs while maintaining the reliability of news article sources
3Adaptability or versatility
If event extraction is performed on multiple media types, then comprehensive event information is achieved, but the process becomes more complex
Solution Approach 1:
The system implements a universal event extraction framework that handles multiple media types (social media, news articles, blogs, forums) through a common architecture. The event schema and extraction algorithms are designed to be media-agnostic, allowing the same system to process diverse sources without requiring separate complex pipelines for each media type
Solution Approach 2:
The system adapts processing parameters based on media type characteristics. Different media sources have their extraction parameters (such as text length thresholds, entity recognition settings, verification intensity) dynamically adjusted according to their specific properties, enabling comprehensive multi-media processing while maintaining operational simplicity
4Loss of time
If social media postings are processed, then real-time event detection is possible, but noise and spam reduce accuracy
Solution Approach 1:
The system converts the harmful noise and spam in social media into beneficial signals by using them as verification data points. Multiple postings about the same event, even if individually noisy, collectively strengthen event detection when patterns emerge. Spam and noise are filtered through statistical analysis where frequent patterns indicate real events while random noise cancels out
Data Source
Figure 1
Figure 2
Figure 3a
AI summary
A method of providing cross-media event linking may include: receiving, at a first input of an event coreferencing system, a stream of social media postings, and at a second input, a stream of news articles; generating, by the event coreferencing system, a first set of event representations representing events referenced by the social media postings, and a second set of event representations representing events referenced by the news articles; determining, by the event coreferencing system, that at least one of the social media postings references a same event referenced by at least one of the news articles, the determining including determining at least one similarity using data of at least one of the first set of event representations.