Multimedia Categorization via User Icon Interaction Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video categorization methods on video websites are inefficient due to reliance on limited information, such as video titles and labels, resulting in low accuracy.
Innovation Solution
A method and device that count the numbers of icons input by users for multimedia resources, specifically bullet-screen and comment icons, to determine the category, improving categorization accuracy by considering user interactions and emotions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If video categorization is performed based on titles and labels, then the categorization process is simple, but the categorization accuracy is low
Solution Approach 1:
The patent applies feedback by collecting user interaction data (comments, likes, shares, playback behavior) and using this feedback to continuously optimize video categorization. The system analyzes user feedback patterns to automatically adjust and improve categorization accuracy over time, transforming static title/label-based categorization into a dynamic, feedback-driven process that learns from user behavior.
Solution Approach 2:
The patent introduces user interaction data as an intermediary element between the video content and the categorization system. Instead of directly categorizing based on video metadata alone, the system uses user behavior data (comments, likes, playback patterns) as a mediating layer to infer and determine more accurate video categories, thereby improving precision without requiring direct analysis of complex video content.
2Measurement precision
If only title and label information is used for categorization, then the data processing is efficient, but the categorization accuracy is low
Solution Approach 1:
The patent segments the categorization process into multiple independent analysis dimensions: title/label analysis, user comment analysis, like/share behavior analysis, and playback pattern analysis. Each dimension processes specific types of data independently, then the results are integrated to form a comprehensive categorization decision. This segmentation allows the system to utilize diverse information sources without overwhelming the processing system.
Solution Approach 2:
The patent transitions from traditional single-dimension categorization (based only on video metadata) to multi-dimensional categorization by incorporating user interaction dimensions. The system analyzes videos across multiple dimensions including textual dimensions (titles, labels, comments), behavioral dimensions (likes, shares, playback completion rates), and temporal dimensions (viewing patterns over time), thereby enriching the information basis for accurate categorization.
Data Source
AI summary
The present disclosure relates to a method and a device for categorizing multimedia resources. The method includes: counting the numbers of icons for respective types of icons input by a user for a multimedia resource; and determining a category to which the multimedia resource belongs according to the numbers of the respective types of icons input for the multimedia resource. The method and device for categorizing multimedia resources according to the present disclosure can take respective types of icons input by the user for the multimedia resources into consideration when categorizing the multimedia resources, thereby improving the accuracy of the categorization of the multimedia resources.