Textual Data Analysis for Analogical Question Answering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current automated chat services face challenges in responding to analogical questions due to the need for extensive and costly construction of RDF knowledge bases, resulting in limited coverage and low success rates for answering such questions, and often provide robotic and terse responses that fail to engage users effectively.
Innovation Solution
The system analyzes textual data to identify candidate analogical questions, extracts source substantives, and uses a word vector model to determine target substantives without relying on a pre-built knowledge base, generating answers by embedding target substantives into analogical answer templates, allowing for more natural and human-like responses.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If RDF knowledge base construction is used to answer analogical questions, then answer accuracy is improved, but system complexity and resource consumption increase significantly
Solution Approach 1:
The patent extracts only the necessary relationship information from unstructured text data and stores it in a simplified format, rather than maintaining a complete RDF knowledge base. This extraction approach answers analogical questions by retrieving relevant relationship patterns from text, reducing system complexity while maintaining answer accuracy.
Solution Approach 2:
The patent creates simplified copies of relationship information from complex RDF structures by extracting essential entity-relation patterns from text data. These copied relationship patterns are sufficient for answering analogical questions without requiring the full RDF knowledge base infrastructure.
2Adaptability or versatility
If complete RDF knowledge base construction is performed, then coverage of analogical questions is improved, but time and resource consumption increase
Solution Approach 1:
The patent performs preliminary extraction of relationship patterns from text data and stores them in a searchable format, enabling quick retrieval when answering analogical questions. This preliminary action avoids the need for time-consuming RDF knowledge base construction while maintaining coverage of various question types.
Solution Approach 2:
The patent extracts and stores only the partial relationship information necessary for answering analogical questions, rather than constructing a complete RDF knowledge base with all possible relationships. This partial action approach achieves sufficient coverage for the target application while reducing construction time and resource consumption.
3Reliability
If RDF-based question answering is used, then factual accuracy is improved, but response naturalness deteriorates
Solution Approach 1:
The patent introduces an intermediary processing layer that retrieves factual relationship information from extracted text data and reformats it into natural, conversational responses. This intermediary transforms rigid RDF-style answers into human-like language while preserving factual accuracy, improving response naturalness without sacrificing reliability.
Data Source
AI summary
Analyzing textual data is disclosed, including by: receiving textual data; determining that the textual data is a candidate for analogy analysis based at least in part on at least a portion of the textual data matching an analogical question template; extracting a source substantive from the textual data; using the source substantive to determine a target substantive from a word vector model that is trained on a set of training data; and generating an answer including the target substantive based at least in part on an analogical answer template corresponding to the analogical question template.