Search Query Question Ranking via Canonical Form Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing question and answer forums lack an efficient method to identify and rank relevant questions from search queries, leading to a suboptimal user experience in content delivery.
Innovation Solution
A system and method that filters and ranks previously-searched questions by removing specified words, determining a canonical form, and ranking questions based on search frequency, using a ranking server with an evaluation engine and ranking engine to generate representative questions for Q&A forums.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual evaluation is used to identify well-formed questions from query logs, then question quality can be ensured, but the process is time-consuming and labor-intensive
Solution Approach 1:
The system enables automatic identification and ranking of well-formed questions through algorithmic processing of query logs. The ranking engine autonomously evaluates questions based on predefined criteria (search frequency, canonical form matching, question word presence) without requiring manual human review, thus achieving both time efficiency and consistent quality assessment
Solution Approach 2:
The system transforms the subjective task of manual question evaluation into an objective parameter-based ranking process. By quantifying question characteristics (search frequency counts, canonical form similarity scores, question word detection), the system automatically identifies high-quality questions through measurable parameters rather than subjective human judgment
2Adaptability or versatility
If all previously-searched questions are presented to users, then comprehensive coverage is achieved, but relevance and quality deteriorate due to noise and duplicates
Solution Approach 1:
The system extracts only the most relevant questions from the pool of previously-searched queries by applying filtering criteria. Questions are evaluated based on search frequency thresholds, canonical form matching, and presence of question words, extracting only those that meet quality standards while excluding noise and duplicates
Solution Approach 2:
The system segments the question identification process into distinct evaluation stages: filtering for question words, removing stop words, generating canonical forms, ranking by search frequency, and selecting top candidates. This segmented approach ensures comprehensive coverage while maintaining high relevance through systematic filtering at each stage
3Productivity
If questions are ranked by search frequency alone, then popular questions are prioritized, but less frequent but high-quality questions may be overlooked
Solution Approach 1:
The ranking engine performs multiple functions simultaneously: it counts search frequency, generates canonical forms for deduplication, checks for question words, and applies quality filters. This multi-functional approach ensures that questions are evaluated on multiple dimensions (frequency, uniqueness, format quality) rather than frequency alone, preventing high-quality but less frequent questions from being overlooked
Data Source
AI summary
The present disclosure is directed to a system and method for posting questions from search queries. In some implementations, a method includes identifying a plurality of different questions previously searched. The previously-searched questions each include a word indicating a question. The previously-searched questions are filtered to remove one or more specified words included with the question word. At least a subset of the plurality of previously-searched questions that can be used to generate a canonical form after removing the one or more specified words are determined. The subset of previously-searched questions are ranked based, at least in part, on a frequency of submission of each previously-submitted search query in the subset. A particular one of the previously-submitted search queries in the subset is identified as representative of the subset of previously-submitted search queries based on the ranking.