Search Recommendation System for Vertical Query Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current search engines are inefficient in meeting users' vertical search requirements, as they fail to intuitively list highly relevant vertical search types, leading to increased human-machine interactions and resource usage, while users often need to perform additional clicks to find relevant results.
Innovation Solution
A recommendation search method that records search and click logs to determine weights of vertical search tendencies for query terms, displaying recommended query terms with associated vertical search types and performing searches using corresponding vertical search engines upon user clicks, thereby reducing unnecessary interactions and improving search efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a universal search technology is used to provide comprehensive search results, then the coverage of information types is improved, but the ability to meet vertical search requirements deteriorates
Solution Approach 1:
The patent segments the search system into multiple vertical search engines, each dedicated to a specific type of information (news, videos, images, music, etc.). This segmentation allows each engine to specialize in its domain, improving vertical search accuracy while the collection of engines collectively maintains broad coverage through the unified search interface.
2Adaptability or versatility
If the search engine displays all recommended query terms with multiple vertical search types, then the completeness of search options is improved, but the complexity of the interface deteriorates
Solution Approach 1:
The patent applies local quality by dynamically adjusting the interface display based on the specific query term. Instead of showing all vertical search types uniformly, the system identifies and displays only the most relevant vertical search types for each query, making the interface simple and focused while maintaining completeness of options through contextual relevance.
3Loss of time
If the search engine performs automatic calculation and display of recommended query terms, then the time for user input is reduced, but the number of human-machine interactions increases
Solution Approach 1:
The patent performs preliminary action by automatically calculating and displaying recommended query terms with their corresponding vertical search types before the user submits their search. This allows users to directly click on pre-processed recommended terms that already have the appropriate vertical search engine identified, reducing both input time and the number of subsequent interactions.
4Measurement precision
If the search engine uses statistical data to filter candidate query terms, then the relevance of recommended terms is improved, but the accuracy of vertical search matching deteriorates
Solution Approach 1:
The patent uses feedback mechanisms by analyzing user search behavior data to continuously optimize the matching between recommended query terms and vertical search types. The system learns from user interactions and adjusts the weighting and selection criteria to improve both the relevance of recommended terms and the accuracy of vertical search type matching simultaneously.
Data Source
AI summary
A recommendation search method is provided for a search engine. The method includes recording a search log and a click log for query terms during search processes, and determining weights of tendencies of vertical searches for the query terms based on the recorded search log and click log. The method also includes, upon detecting a query term in a search box, determining recommended query terms relating to the query term in the search box, searching for the weights of tendencies of the vertical searches for each of the recommended query terms and determining a tendentious vertical search type for each of the recommended query terms based on the weights of tendencies of the vertical searches, and displaying each of the recommended query terms and links of the tendentious vertical search types for the recommended query terms. Further, the method includes, upon detecting a click on a link of one of the tendentious vertical search types corresponding to one of the recommended query terms, performing a search based on the one of the recommended query terms by using a vertical search engine corresponding to the one of the tendentious vertical search type, and displaying search results.