Conversation Graph Message Relevance Identification

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

In social networking services, large conversations with numerous messages overwhelm users, making it difficult to identify relevant information within the conversation graph.

Innovation Solution

A method to identify and mark relevant messages in a conversation graph by selecting a context message, marking authors and parent messages, and iteratively marking additional messages based on criteria such as author credibility and engagement, ultimately providing a list of relevant messages to users.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Loss of information

If all messages in a conversation are displayed to the user, then the user can view the complete conversation history, but the number of messages displayed becomes very large and overwhelms the user

Engineering Contradiction:

Improvecompleteness of conversation informationVSAvoiduser experience and information overload

Core Design Contradiction:

Loss of informationVSEase of operation

Solution Approach 1:

The conversation graph is segmented into relevant and irrelevant messages through iterative pruning. The system divides the complete conversation into meaningful segments by identifying context messages and their related authors, then separating these from the rest of the conversation data that would otherwise overwhelm the user.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The system extracts only the necessary subset of messages from the complete conversation graph. By identifying context messages and iteratively finding messages from related authors and their parent messages, the system takes out and removes irrelevant messages, leaving only the essential conversation elements for user display.

Inventive Principle:

Principle #2Taking out (Extraction)

2Measurement precision

If the system marks messages as relevant based on multiple criteria including author credibility and engagement, then the relevance accuracy improves, but the processing complexity increases

Engineering Contradiction:

Improvemessage relevance accuracyVSAvoidprocessing algorithm complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The system performs preliminary actions by pre-identifying context messages and their authors before the main relevance determination process. By starting with known context messages and their authors, the system establishes a foundation for iterative expansion, reducing the overall processing complexity while maintaining accuracy.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system uses feedback through iterative pruning where each iteration refines the set of relevant messages based on the previous iteration's results. The process continues until convergence, with each cycle improving relevance accuracy by incorporating feedback from identified authors and their message patterns.

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS10229218B1Identifying relevant messages in a conversation graph

Publication Date: 2019.03.12 X CORP

AI summary

A method and system for identifying relevant messages in a conversation graph includes identifying a context message within the conversation graph. The context message is marked as relevant and additional messages are marked as relevant based on various criteria. A plurality of authors of messages marked as relevant is identified and a plurality of unmarked messages authored by the plurality of authors is marked as relevant. In addition, a plurality of parent messages of messages marked as relevant are identified and marked as relevant. A list comprising messages marked as relevant is then provided to a user. Multiple steps can be repeated until convergence after which the list can be provided to the user.