Webpage Type Identification via URL Rule Matching and ML

The patent segments the classification task into two distinct stages: rule-based filtering (using URL patterns, domain names, and path structures) and machine learning classification (using SVM or other classifiers). This segmentation allows the system to leverage the accuracy of manual classification rules for obvious cases while using automated ML for ambiguous cases, thereby improving both accuracy and scalability simultaneously.

The patent introduces an intermediary mechanism - the rule-based filtering system - that acts as a gateway between the incoming URL and the ML classifier. This intermediary pre-processes URLs by applying classification rules first, only passing uncertain cases to the ML model, thus reducing computational overhead while maintaining high accuracy.

The patent applies preliminary action by performing rule-based classification before invoking the computationally intensive ML classifier. The rule-based system quickly filters and classifies URLs that match known patterns (news, shopping, entertainment, etc.), only passing ambiguous cases to the ML model. This preliminary filtering significantly reduces the number of URLs requiring ML processing, thereby reducing overall computation time while maintaining broad field coverage.

The patent implements partial action by applying the full ML classification process only to the subset of URLs that fail to match rule-based criteria. Instead of applying ML to all URLs, the system performs partial classification - using simple rules for clear cases and reserving complex ML processing for ambiguous cases, thus optimizing computation time while maintaining accuracy.

The patent merges two classification approaches - rule-based classification and machine learning classification - into a unified hybrid system. The rule-based component provides fast processing for known patterns, while the ML component supplements it with learned patterns from training data, handling unknown or ambiguous cases. This merging allows the system to achieve both high processing speed and high accuracy simultaneously.

The patent creates a universal classification system that can handle both known and unknown webpage types. The rule-based portion handles conventional, well-documented URL patterns, while the ML classifier provides universal pattern recognition that adapts to various fields and unknown types. This multi-functional approach ensures the system maintains high speed for common cases while achieving high accuracy across diverse and unknown fields.