Query-Interest Mapping for Accurate Visual Content Retrieval
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Online services struggle to effectively respond to text-based queries with visual content, as typical queries are brief and often overloaded, making it challenging to determine the intended context and relevance of visual content items.
Innovation Solution
A system that expands text-based queries by aggregating terms from frequently engaged visual content items, maps the expanded query to an interest taxonomy using a trained machine learning model, and selects relevant content items based on predicted scores for interest nodes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If the system uses brief text-based queries to search for visual content, then the query input is simple and fast, but the relevance and accuracy of retrieved visual content deteriorates
Solution Approach 1:
The system performs preliminary query expansion by aggregating terms from frequently engaged visual content items before the actual search. This preliminary action enriches the brief query with additional context and terms, enabling accurate retrieval without requiring users to input detailed queries.
Solution Approach 2:
The system introduces an intermediary mechanism that maps expanded queries to an interest taxonomy using a trained machine learning model. This intermediary layer translates the enriched query terms into standardized interest nodes, bridging the gap between brief user queries and precise visual content retrieval.
2Measurement precision
If the system aggregates terms from frequently engaged visual content items to expand queries, then the content retrieval accuracy improves, but the processing time and system complexity increases
Solution Approach 1:
The system pre-computes and stores aggregated terms from frequently engaged visual content items in advance. This preliminary preparation allows the system to quickly retrieve and utilize expanded query terms without performing time-consuming aggregation during actual query processing.
Solution Approach 2:
The system replaces complex real-time query expansion mechanisms with a pre-trained machine learning model that maps expanded queries to interest taxonomy. This substitution reduces processing time by using pre-computed mappings instead of performing complex aggregation and ranking operations during query execution.
3Reliability
If the system uses a trained machine learning model to map expanded queries to interest taxonomy, then the relevance of returned content improves, but the device complexity and computational resources increase
Solution Approach 1:
The trained machine learning model serves multiple functions: it maps expanded queries to interest taxonomy nodes, ranks content items based on predicted scores, and handles various query types uniformly. This multi-functionality reduces overall system complexity by consolidating multiple processing steps into a single versatile model.
Solution Approach 2:
The system changes the parameter representation from raw query terms to standardized interest taxonomy nodes through the machine learning model. This parameter transformation simplifies the matching process between queries and content, reducing computational complexity while improving relevance through the learned relationships between terms and visual content characteristics.
Data Source
AI summary
Systems and methods for identifying relevant content within a corpus of visual content items in response to a user's text-based query are presented. In response to a text-based query, the query is mapped to a most-engaged content item of the corpus of visual content items included in responses to the query from a plurality of users. At least one text-based term associated with the most-engaged content item is identified and combined with the query from an expanded query. The expanded query is mapped to an interest node of an interest taxonomy and content items associated with the mapped interest node are identified. At least some of the content items associated with the mapped interest node are selected and returned as response content to the received query.