Synthetic Query Generation via Document Structure Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional search engines struggle to identify and suggest well-performing search queries that are not variations of existing well-performing queries, leading to suboptimal search results when users enter vague or ambiguous queries.
Innovation Solution
The system generates synthetic queries by analyzing structured documents and seed queries, using embedded coding fragments and generative rules to create query templates that can be applied to other documents, measuring performance, and storing high-performing synthetic queries for future use.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If the search engine uses conventional query suggestion methods based on existing well-performing queries, then the system can leverage historical data to suggest queries, but it cannot identify new types of well-performing queries that are not variations of existing queries
Solution Approach 1:
The system performs preliminary analysis of document structures and content before generating query suggestions. By pre-processing documents to extract structural patterns and semantic information, the system prepares a rich feature set that enables discovery of new query types beyond simple variations of existing queries.
Solution Approach 2:
The patent introduces an intermediary representation layer that maps document structures to query templates through a structured intermediary format. This intermediary mechanism translates document content and structure into generative query templates, enabling the system to discover new query types by combining and transforming intermediary representations rather than directly manipulating raw query data.
2Ease of operation
If the search engine returns results for vague or ambiguous queries, then users can access information, but the search results may not be highly ranked or presented on the first page
Solution Approach 1:
The system performs preliminary query refinement by analyzing the vague or ambiguous query and generating multiple candidate queries with different levels of specificity. These candidate queries are pre-processed and ranked before the actual search execution, ensuring that refined queries with better ranking potential are used to retrieve results.
Solution Approach 2:
The system dynamically adjusts query formulation based on the vagueness level of the initial query. For vague queries, the system generates multiple query variations with different specificity levels and uses these dynamic query formulations to retrieve and rank results, allowing the search system to adapt its behavior based on query characteristics.
3Ease of operation
If the search engine relies on user-provided queries, then the system can directly address user needs, but it cannot generate queries that users might not have thought to ask
Solution Approach 1:
The system performs self-service query generation by automatically analyzing document structures and generating relevant queries without requiring user input for query formulation. The system uses its own internal mechanisms to extract query templates from document metadata and content, enabling it to generate queries that users might not have thought to ask.
Solution Approach 2:
The patent creates a universal query generation mechanism that can operate independently of specific user inputs. The system uses a general-purpose template generation approach that can create queries for various types of documents and search scenarios, making the query generation capability universally applicable across different search contexts without requiring domain-specific programming.
Data Source
AI summary
Methods, systems, and apparatus, including computer program products, for generating synthetic queries using seed queries and structural similarity between documents are described. In one aspect, a method includes identifying embedded coding fragments (e.g., HTML tag) from a structured document and a seed query; generating one or more query templates, each query template corresponding to at least one coding fragment, the query template including a generative rule to be used in generating candidate synthetic queries; generating the candidate synthetic queries by applying the query templates to other documents that are hosted on the same web site as the document; identifying terms that match structure of the query templates as candidate synthetic queries; measuring a performance for each of the candidate synthetic queries; and designating as synthetic queries the candidate synthetic queries that have performance measurements exceeding a performance threshold.