Social Media Fake News Detection via Propagation Pathways
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for classifying social media messages into categories like real news or fake news face challenges due to sparse and noisy content, requiring more CPU processing and memory resources, and are inefficient in filtering out unwanted messages.
Innovation Solution
The approach involves obtaining user embeddings in a social media network based on network connectivity, constructing a sequence classifier using LSTM-RNN to model propagation pathways, and predicting class labels associated with messages, thereby filtering out fake news without relying on content information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If content-based approaches are used to classify social media messages, then classification accuracy can be maintained, but CPU processing requirements and memory resources increase significantly
Solution Approach 1:
The patent extracts and utilizes only the necessary propagation pathway information from the social media network, rather than processing entire message contents. By focusing on the sequence of users who spread the message and their network relationships, the system achieves classification without heavy content-based CPU processing
Solution Approach 2:
The patent replaces traditional content-based mechanical text processing with a network-based classification mechanism. Instead of analyzing message text through CPU-intensive algorithms, the system uses propagation pathways and user embeddings to classify messages, significantly reducing computational overhead
2Reliability
If content-based approaches are used to filter social media messages, then classification can be performed, but processing time increases
Solution Approach 1:
The patent performs preliminary action by pre-computing user embeddings based on network connectivity and propagation patterns. These pre-computed embeddings are stored and reused for rapid classification of new messages, eliminating the need for time-consuming content analysis each time a message needs to be classified
Solution Approach 2:
The patent uses propagation pathway patterns as templates to classify new messages. By copying and matching propagation patterns rather than analyzing message content, the system achieves fast classification with minimal processing time
3Measurement precision
If prior art methods are used to detect fake news, then some classification can be achieved, but they fail when content is manipulated to look like real news
Solution Approach 1:
The patent inverts the traditional approach by not analyzing message content to detect fake news, but rather analyzing the propagation pathway and user network patterns. This inversion makes the system immune to content manipulation since fake news with manipulated content but abnormal propagation patterns can still be detected
Solution Approach 2:
The patent introduces propagation pathways as an intermediary between message content and classification. Instead of directly analyzing potentially manipulated content, the system uses the propagation pathway as an intermediate indicator that reveals the true nature of the message, making detection robust against content manipulation
Data Source
AI summary
Messages are transmitted in a social media network. Embeddings of social media network users in the social media network are inferred. Propagation pathways over which the plurality of messages are transmitted through the social media network are classified. Action is taken on one or more of the messages that are transmitted through the social media network, based on the classification of the propagation pathways over which the messages are transmitted through the social media network and the inferred embeddings of the social media network users.