Query Generation Using Heuristic Refinement and Deep Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional approaches to improving user recommendations in online data services, such as news services, face challenges in ensuring high-quality event summaries due to inconsistent user engagement and reliance on popular user search queries that may miss essential event facts.
Innovation Solution
The use of heuristic refinement and a deep learning model to identify high-quality user search queries by filtering queries based on user interactions, determining candidate queries through heuristic and model-based processes, and selecting queries based on aggregated token frequency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional keyword/key phrase extraction is used to identify common key terms in events, then the system can process text data, but it relies heavily on an adequate amount of text data which may not be available
Solution Approach 1:
The patent replaces conventional keyword extraction methods with a deep learning model that uses neural networks to identify key terms and generate event summaries. This substitution allows the system to achieve high-quality summaries without relying heavily on large amounts of text data, as the deep learning model can extract meaningful patterns from limited data through learned representations
Solution Approach 2:
The patent changes the approach from traditional text processing parameters to deep learning model parameters, including token embeddings, attention weights, and neural network weights. This parameter transformation enables the system to maintain high measurement precision (summary quality) even when the quantity of text data is insufficient for conventional methods
2Productivity
If popular user search queries are used as primary sources for recommendations, then the system can leverage user interest, but the queries may miss essential event facts reducing summary quality
Solution Approach 1:
The patent merges multiple query generation approaches by combining deep learning model outputs with heuristic refinement processes. This combination integrates the efficiency of popular query patterns with the precision of fact-based event analysis, producing recommendations that maintain both user interest alignment and essential event fact accuracy
Solution Approach 2:
The patent introduces an intermediary refinement process that acts as a mediator between raw user search queries and final recommendations. This intermediary layer filters and enhances queries to ensure they capture essential event facts while maintaining their relevance to user interest, thereby improving summary quality without sacrificing recommendation efficiency
3Reliability
If user views and clicks are used to determine popular content for recommendations, then the system can measure user engagement, but user involvement is typically inconsistent impacting recommendation quality
Solution Approach 1:
The patent implements a self-service mechanism where the deep learning model automatically adapts to varying user engagement patterns without requiring complex external intervention. The model learns from available interaction data and adjusts its query generation and recommendation strategies autonomously, maintaining reliable and consistent recommendations even when user involvement fluctuates, while avoiding excessive system complexity
Data Source
AI summary
Systems and methods are disclosed for using heuristic refinement and a deep learning model to identify at least one high-quality user search query. One method comprises filtering one or more queries, determining at least one candidate query based on the filtered queries, the determining including: determining heuristic candidate queries by applying heuristic processes to the filtered queries, generating model candidate queries based on one or more deep learning model processes, and selecting the candidate query from the heuristic candidate queries or the model candidate queries based on a corresponding aggregated token frequency, and displaying the candidate query.