Personalized NLS Query Suggestions Using Co-occurrence Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional search engines in multi-tenant systems face challenges in providing both efficient and relevant natural language search results, as they are either computationally intensive, leading to slow performance, or they provide poorly ranked results due to not processing all relevant information, which affects user experience and relevance for atypical users.
Innovation Solution
The implementation of a per-user co-occurrence model and a modified suggestion pipeline that learns usage patterns from query logs to generate personalized natural language search suggestions, ensuring scalability and privacy compliance within a multi-tenant architecture.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If search engines process all relevant information to rank search results, then search result quality is improved, but computational resources required increase significantly
Solution Approach 1:
The search ranking process is segmented into two distinct stages: a filtering stage that uses co-occurrence models to identify and remove irrelevant results quickly, and a ranking stage that applies full relevance analysis only to the remaining candidate results. This segmentation allows the system to achieve high search result quality without requiring computational resources for all results, resolving the contradiction between processing all relevant information and the computational cost involved.
Solution Approach 2:
The invention extracts and removes irrelevant search results early in the processing pipeline using co-occurrence models trained on user behavior patterns. By taking out irrelevant results before they reach the full ranking stage, the system reduces the computational burden while maintaining high search result quality among the remaining relevant results.
2Speed
If search engines use conventional ranking techniques, then speed is improved, but search result relevance deteriorates
Solution Approach 1:
The system performs preliminary action by pre-training co-occurrence models on historical user search data before actual search queries are processed. These models capture user-specific information preferences and are ready to immediately filter results during search operations. This preliminary training enables the system to achieve both fast search speeds and high result relevance without requiring intensive computational resources during the actual search process.
Solution Approach 2:
The co-occurrence model serves as an intermediary between the user's search query and the full search index. It acts as a mediator that uses learned patterns from historical data to pre-filter and prioritize results, enabling the system to return relevant results quickly without performing exhaustive analysis on all data, thus resolving the speed-relevance contradiction.
3Productivity
If search engines provide common search suggestions, then search efficiency is improved for typical users, but relevance deteriorates for atypical users
Solution Approach 1:
The invention applies local quality by transitioning from uniform common search suggestions to personalized suggestions tailored to each user's specific behavior patterns. The co-occurrence models analyze individual user search history and preferences to generate locally optimized suggestions for each user, rather than applying a one-size-fits-all approach. This enables high search efficiency and suggestion relevance for atypical users by customizing suggestions to their unique needs.
Solution Approach 2:
The search suggestion system becomes dynamic by adapting to each user's individual search behavior patterns over time. The co-occurrence models continuously learn from user interactions and update personalized suggestions accordingly. This dynamic adaptation allows the system to maintain high search efficiency while improving suggestion relevance for atypical users whose search patterns differ from the average user behavior.
Data Source
AI summary
A method and apparatus for generating personalized suggestions for natural language search queries, where the method includes receiving a natural language query input from a user, obtaining set of suggestions for the natural language query, identifying a set of concepts in the set of suggestions, applying co-occurrence model to the set of concepts to obtain user specific concept probabilities and vocabulary probabilities, computing suggestion scores for at least one suggestion in the set of suggestions based on the user specific concept probabilities and vocabulary probabilities, and ranking the set of suggestions based on suggestion scores.