Natural Language Query Clustering via Event Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current machine learning systems for natural language queries are limited in their ability to cluster queries based on significant events, as they primarily rely on keywords within the query and do not effectively identify deeper relationships between entities, leading to incomplete and irrelevant search results.
Innovation Solution
A method that uses a knowledge graph to extract deep-level entities and events by calculating mutual information scores, allowing for the identification of significant events and their relationships, and iteratively analyzes these to cluster natural language queries based on commonalities in returned events.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If keyword-based search is used, then search speed is improved, but search completeness deteriorates
Solution Approach 1:
The patent transitions from surface-level keyword matching to deep-level event extraction by adding temporal and contextual dimensions. The system analyzes events, time periods, and entity relationships beyond simple keyword occurrence, enabling comprehensive search without sacrificing speed through efficient event-based indexing.
Solution Approach 2:
The patent introduces event extraction as an intermediary layer between keyword search and result retrieval. Instead of directly matching keywords to documents, the system first extracts significant events from queries and documents, then matches events to find relevant results, preserving both speed and completeness.
2Loss of information
If deep-level entity analysis is performed, then information completeness is improved, but computational complexity deteriorates
Solution Approach 1:
The patent extracts only the most significant events and entities from documents and queries, rather than analyzing all content. By focusing on key events that capture the essence of the information, the system achieves deep-level analysis without processing the entire document corpus, reducing computational complexity while maintaining information completeness.
Solution Approach 2:
The patent changes the analysis parameters from comprehensive text processing to event-based analysis with specific temporal and contextual parameters. This parameter transformation enables deep-level understanding by focusing on structured event attributes (time, location, participants) rather than unstructured text, simplifying the computational task.
3Reliability
If event-based clustering is implemented, then query relevance is improved, but processing time deteriorates
Solution Approach 1:
The patent performs preliminary event extraction and clustering of documents before query processing. By pre-organizing documents into event-based clusters and extracting key event features in advance, the system can quickly match queries to relevant clusters without performing deep analysis during actual search operations, improving query relevance while reducing processing time.
Data Source
AI summary
Clustering a set of natural language queries NLQs based on a set of significant events retrieved from a corpus stored in a computer system is described. A set of NLQs is used by a search engine for searching a selected corpus to retrieve respective sets of significant events. The set of NLQs is clustered into a plurality of NLQ clusters according to a number of common significant events being returned by the search engine for respective members of an NLQ cluster.