Hybrid QA System Using Retrieval Feedback for Deep Model Training
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current automatic question-answering systems face challenges in generating accurate answers to complex natural language questions due to reliance on proper question formulation and the need for large training data, leading to inefficiencies and increased operational costs in consumer support.
Innovation Solution
A hybrid question-answering system that combines a retrieval QA application with a deep QA application, where answers from the retrieval system are used to train the deep QA model, improving its accuracy over time by feeding only high-confidence answers as training data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a deep QA application using neural networks is used to generate accurate answers, then answer accuracy is improved, but the need for extensive training data increases
Solution Approach 1:
The patent combines a retrieval-based QA system with a deep learning-based QA system into a hybrid architecture. The retrieval system provides initial answers that are then used to train the deep learning model, allowing the system to achieve high accuracy while reducing the need for extensive manual training data through automated knowledge acquisition
Solution Approach 2:
The system implements a feedback loop where the retrieval QA application generates answers that are fed back as training data to improve the deep QA application. This continuous feedback mechanism allows the deep learning model to learn from real queries and answers, improving accuracy over time without requiring large volumes of pre-collected training data
2Quantity of substance
If a retrieval QA application is used to generate answers, then less training data is needed, but answer accuracy for complex questions deteriorates
Solution Approach 1:
The patent merges the strengths of retrieval-based QA (low training data requirements) with deep learning-based QA (high accuracy for complex questions) in a hybrid system. The retrieval system handles initial queries efficiently while the deep learning component processes complex questions that require deeper understanding, achieving both data efficiency and high accuracy
3Measurement precision
If QA systems rely on proper question formulation to generate accurate answers, then answer accuracy is improved, but the system becomes less adaptable to poorly formulated questions
Solution Approach 1:
The patent introduces an intermediary retrieval system that acts as a bridge between the user's question and the deep QA application. This intermediary layer can handle poorly formulated questions by retrieving relevant information before passing it to the deep learning model, which then generates accurate answers even when the input question is not optimally formulated
Data Source
AI summary
Methods, apparatus, systems, computing devices, computing entities, and/or the like for a hybrid question-answering (QA) application are provided. The hybrid QA application includes both a retrieval QA application and a deep QA application, and uses feedback from the retrieval QA application to improve the deep QA application.