Category Aspect Mining for Search Relevance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional information retrieval systems fail to provide authors with relevant category aspect information that users seek, leading to less accurate search results and increased computational resources used for navigation.
Innovation Solution
A system that mines historical user behavior data to determine demand scores for category aspects, sorting them to identify the most relevant aspect names and values, which are then used to guide publishers in creating more relevant listings and promote them in search results.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional information retrieval systems are used, then authors can publish descriptions, but the descriptions are not relevant to what users seek, resulting in inaccurate search results
Solution Approach 1:
The system performs preliminary mining of category aspect information from historical user behavior data before search queries are executed. By pre-identifying relevant aspects and their demand scores, the system prepares structured information that can be immediately applied to improve search result accuracy without adding computational overhead during actual search operations.
Solution Approach 2:
The system utilizes feedback from historical user behavior data to continuously identify and refine category aspect information. By analyzing past user interactions, clicks, and search patterns, the system determines which aspects are most valuable to users and uses this feedback to improve the relevance and accuracy of search results over time.
2Ease of operation
If authors provide descriptions without guidance, then publishing is simple, but the descriptions lack relevance to user search intent
Solution Approach 1:
The system provides self-service by automatically generating and providing guidance information to authors about relevant category aspects for their publications. Rather than requiring authors to manually research what information users need, the system autonomously mines this information from user behavior data and presents it as guidance, maintaining publishing simplicity while improving description relevance.
Solution Approach 2:
The system acts as an intermediary between user behavior data and authors. It extracts category aspect information from historical user interactions and translates it into actionable guidance for authors, bridging the gap between what users seek and what authors provide without requiring direct author involvement in the analysis process.
3Reliability
If users perform additional searching and navigation to find relevant information, then comprehensive results can be found, but computational resources such as processor cycles, network traffic, and power consumption increase
Solution Approach 1:
The system performs preliminary organization of information by category and aspect before users need to search. By pre-structuring data according to mined category aspects and demand scores, the system enables more direct and accurate search results, reducing the need for users to perform multiple navigation steps and thereby decreasing computational resource consumption during search operations.
Data Source
AI summary
In various example embodiments, systems and methods for providing category aspect information by mining historical data is provided. In example embodiments, a table comprising joined data is accessed. The table includes historical data that comprises user behavior data based on actions performed with past queries by users, data describing publication, and a determined category for each publication. Demand scores based on the joined data are determined. The determined demand scores are used to determine most relevant aspect name and aspect value pairs for a category. Publications having at least one of the most relevant aspect name and aspect value pairs are displayed visually distinguished from less relevant publications that exclude the most relevant aspect name and aspect value pairs.