Topical Phrase Identification in Browsing Sequences
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Internet users frequently revisit pages, making it difficult to re-find previously visited sites without effective navigation aids, as existing technologies do not adequately utilize the context of browsing sequences to identify topically-related phrases.
Innovation Solution
The technique extracts topically-related phrases from a user's browsing sequence by analyzing pages individually and those visited before and after, filtering them based on a pre-populated topic list, and clustering to find the most relevant ones, with re-ranking based on user browsing history.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If phrases are extracted from individual pages in isolation, then the extraction process is simple and fast, but the accuracy of identifying topically-related phrases is poor
Solution Approach 1:
The browsing sequence is divided into multiple pages, and each page is processed individually to extract candidate phrases. However, these individual extractions are then integrated through co-clustering of pages sharing candidate phrases, resolving the contradiction between simple individual processing and accurate topical identification.
Solution Approach 2:
Pages are co-clustered based on shared candidate phrases to identify topically-related phrases. This merging of page-level information enables accurate topical phrase identification while maintaining the simplicity of individual page processing through the co-clustering mechanism.
2Productivity
If all phrases from browsing sequence pages are collected, then the quantity of candidate phrases is large, but filtering and identifying relevant phrases becomes time-consuming
Solution Approach 1:
A prescribed phrase list is pre-populated with candidate phrases before processing the browsing sequence. This preliminary preparation enables efficient filtering of extracted phrases against the pre-existing list, significantly reducing the time required to identify relevant phrases while maintaining comprehensive coverage.
Solution Approach 2:
Candidate phrases are extracted from pages and then filtered by selecting only those present in the pre-populated phrase list. This extraction followed by selective filtering efficiently reduces the quantity of phrases to process while maintaining productivity through the optimized filtering mechanism.
3Measurement precision
If phrases are analyzed without considering browsing context, then the analysis process is fast and simple, but the relevance of identified phrases to user intent is reduced
Solution Approach 1:
The browsing sequence is segmented into individual pages with their temporal contexts preserved. This segmentation allows fast individual page processing while the sequential arrangement maintains browsing context, enabling accurate relevance identification without excessive time consumption.
Solution Approach 2:
The browsing sequence is pre-organized with temporal ordering of pages before phrase extraction. This preliminary structuring preserves contextual information that enhances phrase relevance to user intent while maintaining efficient processing speeds through the pre-established temporal framework.
Data Source
AI summary
Browsing sequence phrase identification technique embodiments are presented that generally extract topically-related phrases from the pages visited by a user in a browsing session. The topically-related phrases can be used for a variety of purposes, including aiding a user in re-finding previously visited sites. This phrase identification task is performed by considering not just the pages of a user's browsing sequence individually, but also pages visited immediately before and immediately after each page. In this way, phrases found in a page can be analyzed in the context in which the page was viewed, rather than in isolation. The identified phrases are further filtered by picking those that appear on a pre-populated topic list, and then clustering to find the most informative ones.