Search Intent Categorization via Category Embeddings
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing search technologies consume excessive computing resources due to repetitive user queries and incorrect categorization of search intent, leading to inefficient search results and increased network latency and storage device I/O operations.
Innovation Solution
A search system generates category embeddings that encode hierarchical and co-occurring relationships between categories, allowing for improved query categorization by selecting core intent categories, thereby reducing repetitive queries and optimizing resource usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional query understanding methods are used to identify search intent, then search results can be returned, but excessive computing resources are consumed due to repetitive queries and incorrect categorization
Solution Approach 1:
The system performs preliminary actions by pre-processing search information to generate category embeddings that encode hierarchical relationships and co-occurring categories before queries are submitted. Category clusters are formed in advance, and core intent categories are pre-identified for each cluster, so that when a query arrives, the system can quickly match it to the appropriate pre-computed category rather than performing complex analysis in real-time.
Solution Approach 2:
The system creates simplified copies of category information in the form of category embeddings and category clusters. These embeddings are compressed representations that capture the essential hierarchical and co-occurring relationships between categories, allowing the system to work with these lightweight copies instead of the full, complex category taxonomy during query processing.
2Productivity
If simple category filtering is used for search results, then processing is fast, but search intent may be incorrectly identified leading to irrelevant results
Solution Approach 1:
The system introduces category embeddings as an intermediary layer between the raw query and the final category assignment. These embeddings serve as mediators that capture nuanced relationships between categories, allowing the system to make more accurate categorization decisions while maintaining efficient processing through the use of pre-computed embeddings and clustering.
3Ease of operation
If repetitive user queries are allowed, then users can refine their search, but network latency and storage device I/O operations increase
Solution Approach 1:
The system performs preliminary categorization using the pre-computed category embeddings and clusters to quickly identify the most likely intent categories before any network traversal occurs. This preliminary action filters and prioritizes search results based on the categorized intent, reducing the need for users to submit multiple refined queries and thereby reducing network latency and repetitive I/O operations.
Data Source
AI summary
A search system performs item retrieval using search query categorization that matches query intent. Category embeddings are generated for categories based on hierarchical data and search information. For instance, the category embeddings can be generated using information regarding hierarchical relationships between the categories, co-occurring relationships between categories identified from search information, and initial embeddings that encode query-related information for each category. Category clusters can be formed using the category embeddings. When a search query is received, one or more categories are identified from a category cluster and used for selecting search results for the search query.