Personalized Query Suggestions via User Profile Embedding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing online connection network systems provide the same query suggestions to different users with the same initial query, despite varying search intents based on their professional profiles, location, and behavior, leading to unsatisfactory results.
Innovation Solution
Extending the neural sequence-to-sequence model to incorporate personalized features such as profile industry, language, geographic location, and user behavior, allowing for localized and intent-specific query suggestions by concatenating personalization features with word embeddings or hidden layers in the model.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If the same query suggestions are provided to all users with the same initial query, then the system complexity is reduced and ease of operation is improved, but the relevance and usefulness of search suggestions deteriorates due to ignoring individual search intents
Solution Approach 1:
The system pre-collects and stores user profile information including professional background, industry, location, and behavior data before search queries are submitted. This preliminary preparation enables the model to access relevant personalization features when generating query suggestions, resolving the contradiction by having information ready in advance rather than requiring complex real-time processing
Solution Approach 2:
The patent extends the traditional sequence-to-sequence model by adding a new dimension for personalization features. Instead of only processing the query sequence, the model now incorporates user-specific features as additional input dimensions, enabling personalized suggestions without fundamentally redesigning the core architecture
2Loss of information
If personalized features are incorporated into the sequence-to-sequence model to generate intent-specific query suggestions, then the relevance of search suggestions is improved, but the model complexity and difficulty of implementation increases
Solution Approach 1:
The patent merges the personalization feature processing with the existing sequence-to-sequence model architecture. User profile features are concatenated with query embeddings or integrated into hidden layers, combining multiple information sources within a unified model structure rather than creating separate systems
Solution Approach 2:
The extended model serves multiple functions: it processes standard query suggestions while simultaneously incorporating personalized user features. The same base model architecture handles both generic and personalized suggestion generation, making the system multi-functional without requiring entirely separate models
Data Source
AI summary
Machine learning based method for generating personalized query suggestions is described. Different users may have different search intent even when they are inputting the same search query. The technical problem of personalizing search query suggestions produced by a machine learning model is addressed by extending the sequence to sequence machine learning model framework to be able to take into consideration additional, personalized features of the user, such as, e.g., profile industry, language, geographic location, etc. This methodology includes an offline model training framework as well as an online serving framework.