Live Comment Grouping by Semantic Similarity for Faster Reading
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In live-streaming scenarios, streamers often ignore repeated comments due to rapid refresh rates, leading to reduced reading efficiency as they are overwhelmed by numerous similar comments.
Innovation Solution
A method and device that combines semantically similar comments using a pre-trained neural network model, displaying them as a group to reduce space occupation and prioritize other comments, utilizing a first and second comment area on the interface.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If real-time comments are displayed individually with high refresh rates, then the completeness of comment information is improved, but the reading efficiency deteriorates due to overwhelming repeated comments
Solution Approach 1:
The patent merges semantically similar comments into unified display units. The semantic similarity determination module calculates similarity between real-time comments and historical comments, and when similarity exceeds a threshold, the comments are merged into a single display unit showing the comment content and count of similar comments. This reduces the number of individual comment displays while preserving information about repeated comments.
2Quantity of substance
If all real-time comments are displayed separately, then the quantity of displayed comments is improved, but the display space efficiency deteriorates due to occupation by repeated comments
Solution Approach 1:
The patent combines multiple semantically similar comments into single display units. The display module shows merged comments with their aggregation count (e.g., '10 people are sending the same message'), thereby reducing the total number of display elements and freeing up display space while maintaining visibility of comment quantity and diversity.
3Measurement precision
If semantic similarity determination is performed for all comment pairs, then the accuracy of comment grouping is improved, but the computational complexity increases
Solution Approach 1:
The patent pre-trains a semantic similarity determination model offline before deployment. The model learns semantic relationships from training data and stores the learned parameters. During real-time operation, the pre-trained model quickly computes similarity between comments using the stored parameters, avoiding the need for complex real-time training and reducing online computational complexity while maintaining grouping accuracy.
Data Source
AI summary
The present disclosure relates to a comment processing method and apparatus, and an electronic device, a computer-readable storage medium and a computer program. The comment processing method includes: acquiring real-time comment information for a live room; determining a similarity between a target real-time comment in the real-time comment information and historical comments in a historical comment set corresponding to the live room; when the similarity between the target real-time comment and a target historical comment is greater than or equal to a first preset threshold value, combining the target real-time comment and the target historical comment, wherein the target historical comment is any historical comment among the historical comments in the historical comment set, and the target real-time comment is any real-time comment in the real-time comment information; and displaying a combined target real-time comment and target historical comment.