Query Completion Ranking via Co-occurrence Probability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing search engines do not effectively consider user interests when providing query completions, often ranking irrelevant suggestions due to lack of personalized data, even when a user's previous queries indicate specific interests.
Innovation Solution
A system that adjusts initial query completion rankings by identifying queries likely to co-occur with a user's previous submissions across multiple user activity sessions, boosting relevant completions and filtering out less relevant ones, using a modified ranking score based on co-occurrence probabilities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If query completions are ranked based on general popularity or frequency, then the ranking process is simple and fast, but the relevance to individual user interests deteriorates
Solution Approach 1:
The system pre-computes and stores co-occurrence probability matrices that map reference queries to likely queries. When a user submits a query, the system retrieves the pre-computed co-occurrence probabilities for that reference query and applies them to rank completions, avoiding real-time computation while maintaining personalized relevance
Solution Approach 2:
The patent introduces co-occurrence probability as an intermediary factor between the reference query and query completions. Instead of directly analyzing user history or using simple frequency counts, the system uses pre-computed co-occurrence probabilities as a mediator to bridge the gap between general query data and personalized recommendations
2Reliability
If the system considers user-specific query history to personalize completions, then relevance to user interests improves, but the complexity of processing and storing user data increases
Solution Approach 1:
Instead of storing and processing detailed user-specific query histories, the system creates a simplified copy of user behavior patterns in the form of co-occurrence probability matrices. These matrices capture the essence of user interests through aggregated data from multiple users, providing personalized recommendations without requiring complex user profile storage and processing
Solution Approach 2:
The system transforms detailed user query history into a different parameter representation - co-occurrence probabilities. By changing the data representation from raw user queries to probability distributions over likely queries, the system reduces processing complexity while maintaining the ability to provide personalized recommendations
3Productivity
If the system uses only matching query prefixes from user history, then the processing is fast and simple, but the ability to consider broader user interests deteriorates
Solution Approach 1:
The patent extends the analysis from the one-dimensional matching of query prefixes to a multi-dimensional exploration of query relationships through co-occurrence probabilities. By considering not just exact prefix matches but also probabilistic relationships between queries, the system operates in a higher dimension of query space, enabling broader interest detection without significant performance penalty
Data Source
AI summary
Methods, systems, and apparatus, including computer programs encoded on computer storage media, for reranking query completions based on activity session data. One of the methods includes receiving a query prefix from a user. Query completions are obtained for the query prefix. One or more likely queries that are likely to co-occur with a reference query in user activity sessions are obtained. If one of the likely queries matches one of the query completions, a modified ranking of the query completions is determined, including boosting a ranking of matching query completions. The modified ranking of the query completions is provided in response to receiving the query prefix.