Question Classification and Feature Mapping for Deep QA Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Deep question answering systems face inefficiencies in processing frequent and similar questions, particularly in real-time environments like financial services and social networking, due to the need for extensive analysis and resource-intensive processing.
Innovation Solution
The system classifies questions, identifies highly relevant features, and stores their mappings to skip unnecessary processing steps, while monitoring evidence for changes that may impact answers, allowing for partial or parallel pipeline execution and notification of users when evidence changes affect responses.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If deep question answering systems perform extensive analysis to ensure accuracy, then answer reliability is improved, but processing time and resource consumption increase
Solution Approach 1:
The system performs preliminary classification of questions into categories and identifies relevant features in advance during training. This preliminary action enables the system to skip unnecessary processing steps when answering similar questions, reducing processing time while maintaining accuracy through pre-computed feature mappings.
Solution Approach 2:
The question answering process is segmented into distinct phases: classification phase, feature identification phase, and answer generation phase. By segmenting the process and using the results of earlier phases to guide later phases, the system avoids redundant analysis and optimizes processing efficiency.
2Reliability
If the system processes all features comprehensively, then answer completeness is improved, but computational resource usage increases
Solution Approach 1:
The system extracts and identifies only the most relevant features for each question category during training. During execution, it uses these pre-identified relevant features to generate answers, skipping analysis of irrelevant features and reducing computational resource usage while maintaining answer completeness.
Solution Approach 2:
Different features are weighted and processed differently based on their relevance to specific question categories. The system applies local quality by focusing computational resources on features that are most important for each particular question type rather than uniformly processing all features.
3Measurement precision
If the system uses extensive analysis pipelines, then answer precision is improved, but system complexity increases
Solution Approach 1:
The complex analysis pipeline is simplified through preliminary classification and feature mapping during training. Once the classification model and feature mappings are pre-computed, the actual answer generation process becomes simpler and faster, reducing operational system complexity while maintaining precision.
Data Source
AI summary
System, method, and computer program product to identify relevant features in a deep question answering system, by classifying a first case received by the deep question answering system, and, while training the deep question answering system to answer the first case, identifying a first feature in the first case, computing a first feature score for the first feature, the first feature score indicating a relevance of the first feature in generating a correct response to the first case, and, identifying the first feature as relevant in answering the classified first case upon determining that the first feature score exceeds a relevance threshold.