Notification Server Using Semantic Clustering for Harmful Site Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems lack an effective method to prevent client terminals from accessing harmful sites, such as those with security risks, gambling, or obscenity, by failing to provide timely and accurate notifications to managers.
Innovation Solution
A notification service server that crawls text from accessed sites, clusters words based on semantic similarity, and verifies if the clusters match pre-selected harmful words, sending a warning message to a manager terminal if a threshold similarity is exceeded, along with encrypting terminal identification information for secure transmission.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If text analysis is performed on accessed sites to detect harmful content, then harmful sites can be identified, but the system lacks accuracy in distinguishing harmful from benign sites
Solution Approach 1:
The patent segments the text analysis process into multiple independent components: crawling unit for text extraction, word search unit for keyword matching, clustering unit for grouping similar words, and verification unit for final validation. This segmentation allows each component to specialize in one aspect of analysis, improving overall detection accuracy while reducing false positives through distributed validation.
Solution Approach 2:
The patent introduces intermediate processing layers between raw text and final harmful site determination. The word clustering acts as an intermediary that groups semantically similar words, and the verification unit serves as an intermediary validation step that cross-checks findings against multiple criteria before triggering a harmful site alert, thereby reducing false positives.
2Reliability
If comprehensive text crawling and analysis is performed on all accessed sites, then harmful content detection capability is improved, but system complexity and processing time increase
Solution Approach 1:
The system divides the complex analysis task into four distinct functional modules: crawling unit, word search unit, clustering unit, and verification unit. Each module handles a specific aspect of the analysis, making the overall system more manageable and maintainable while comprehensively analyzing site content through coordinated operation of these segmented components.
Solution Approach 2:
The patent implements preliminary actions by pre-processing and clustering words before final verification. The clustering unit pre-groups semantically similar words into categories, and the verification unit pre-validates against multiple criteria before making a final determination. This preliminary processing reduces the complexity of the final decision-making process.
3Measurement precision
If word clustering based on semantic similarity is implemented, then detection accuracy is improved, but processing time increases
Solution Approach 1:
The patent performs word clustering as a preliminary action before final verification. By pre-grouping semantically similar words into clusters, the system reduces the search space for the verification unit, which only needs to validate against cluster representatives rather than individual words. This preliminary clustering improves accuracy while minimizing time loss through efficient pre-processing.
Solution Approach 2:
The patent implements partial clustering rather than exhaustive analysis of all possible word relationships. The clustering unit groups words based on semantic similarity thresholds, performing sufficient clustering to improve accuracy without exhaustively analyzing every possible word combination, thus balancing precision with processing efficiency.
Data Source
AI summary
Disclosed are a notification service server and an operating method thereof. The present invention presents a notification service server capable of providing an access notification service to harmful sites and an operating method thereof to support a manager to prevent a client terminal from accessing a harmful site.