Browsing Log Segmentation for Search Goal Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current Internet search engines face challenges in accurately determining user search goals from browsing logs, leading to less relevant search results and ineffective personalization.
Innovation Solution
The system classifies and segments user browsing logs by extracting web page features, generating classifiers based on these features, and analyzing pairs of web pages to determine their similarity in search goals, using methods like pairwise classification and gradient boosted decision trees to identify logical browsing sessions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional search engines store and analyze browsing logs to improve search accuracy, then search result relevance improves, but the complexity of analyzing and interpreting browsing log data increases significantly
Solution Approach 1:
The patent segments browsing logs into distinct logical browsing sessions based on user behavior patterns and temporal characteristics. By dividing the continuous browsing log into discrete sessions, the system can analyze each session independently to determine search goals, reducing the overall complexity of browsing log analysis while improving the precision of search result relevance.
2Measurement precision
If the system performs detailed pairwise classification of web address features to determine search goals, then the accuracy of user intent detection improves, but the computational time and processing resources increase
Solution Approach 1:
The patent performs preliminary classification of web address features into categories (such as news, shopping, social media, etc.) before conducting pairwise comparison. This pre-processing step organizes the data structure and reduces the search space for subsequent analysis, enabling accurate user intent detection while minimizing computational processing time.
3Adaptability or versatility
If the system segments browsing logs into multiple logical sessions with different search goals, then the personalization of search results improves, but the complexity of generating and managing multiple classifiers increases
Solution Approach 1:
The patent develops a universal classifier framework that can handle multiple search goal categories using a consistent classification approach. The same classifier structure and methodology are applied across different types of browsing sessions (news, shopping, social media, etc.), enabling search personalization without requiring separate specialized classifiers for each category, thus reducing classifier management complexity.
Data Source
AI summary
Disclosed are systems, methods and computer program products for classification and segmentation of user's browsing logs based on search goals of the user. In one aspect, the method includes receiving a browsing log of a user; extracting web page features from the browsing log; generating web page classifiers based on the extracted features of the browsing log, wherein different classifiers are related to different search goals of the user; segmenting the browsing log using the one or more classifiers into a plurality of separate logical browsing sessions related to different search goals of the user; and performing an Internet search based on a search query of the user and information about one of the logical browsing sessions.