Answer Ranking in Community Q&A Sites via Analogical Reasoning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Community question-answering sites face challenges in distinguishing high-quality answers from low-quality ones due to a lexical gap between questions and answers, exacerbated by user-generated spam and flippant responses, which complicates the identification of accurate answers.
Innovation Solution
An analogical reasoning-based ranking technique that utilizes previous relevant knowledge to bridge the lexical gap by modeling latent relationships between questions and answers, leveraging textual and non-textual features to identify high-quality answers through a logistic regression model and set-based reasoning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional techniques enrich textual features with non-textual features (answerers' category specialty, questioners' self-evaluation, users' votes), then answer ranking accuracy is improved, but device complexity increases
Solution Approach 1:
The patent segments the feature extraction process into distinct modules: textual feature extraction (questions and answers), non-textual feature extraction (user profiles, votes, timestamps), and linkage prediction modeling. This segmentation allows each component to be optimized independently while working together to resolve the lexical gap between questions and answers, improving ranking accuracy without overwhelming complexity
Solution Approach 2:
The patent introduces an intermediary linkage prediction model that mediates between textual and non-textual features. This model processes both feature types through a unified framework, extracting latent relationships that connect questions to answers while filtering out spam and flippant responses, thereby achieving accurate ranking without requiring all features to be processed simultaneously
2Measurement precision
If the system processes both textual and non-textual features to bridge the lexical gap, then answer quality identification is improved, but processing time increases
Solution Approach 1:
The patent applies preliminary action by pre-processing and storing non-textual features (user profiles, category information, vote patterns) in advance. These pre-computed features are then quickly retrieved and integrated during answer ranking, avoiding time-consuming real-time analysis of all features while maintaining high accuracy in identifying quality answers
Solution Approach 2:
The patent changes the parameter representation by transforming raw non-textual data (user profiles, votes) into normalized feature vectors that capture essential patterns. This parameter transformation enables efficient processing and comparison, reducing computation time while preserving the discriminatory power needed to identify high-quality answers among spam and flippant responses
Data Source
AI summary
In some implementations, a plurality of first questions and corresponding first answers are identified at a community question-answer (CQA) site as a plurality of first question-answer (q-a) pairs. A query thread comprised of a second question and a plurality of candidate second answers is selected for making a determination of answer quality. A set of the first questions that are similar to the second question are identified from the plurality of first questions. First linking features between the identified set of first questions and their corresponding first answers are used for determining an analogy with second linking features between the second question and candidate answers for ranking the candidate answers.