Question Answering Device Dynamic Machine Reading Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing question and answer techniques using neural networks often fail to provide accurate answers for questions requiring explanatory responses, as they tend to output only partial answers or not retrieve necessary information, leading to low accuracy due to the inability to determine the necessity of machine reading based on question types.
Innovation Solution
A question and answer device that includes a question determination unit to classify question types and determine whether machine reading is necessary, setting passage data as the answer when not necessary and extracting answer text data via machine reading when necessary, thereby improving answer accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If extractive neural machine reading is used to answer questions, then answer accuracy for simple questions is improved, but answer accuracy for explanatory questions deteriorates because the model outputs only partial answers
Solution Approach 1:
The system dynamically adjusts its behavior based on question type classification. For simple questions, it uses extractive machine reading to provide precise answers. For explanatory questions, it switches to generating comprehensive answers from passages. This dynamic adaptation resolves the contradiction by making the system's response style dependent on the question type rather than using a fixed approach.
Solution Approach 2:
The system changes the parameter of answer generation mode based on question characteristics. By classifying questions into simple vs. explanatory types, it adjusts the processing parameters to either extract specific answers or generate comprehensive explanations, thereby optimizing accuracy for each question type while maintaining versatility across both types.
2Measurement precision
If machine reading is performed on all passages, then answer completeness is improved, but processing time increases
Solution Approach 1:
The system segments the processing workflow into two stages: first classifying the question type, then selectively processing passages based on the classification. For simple questions, it processes only the necessary passages for extraction. For explanatory questions, it processes passages more thoroughly. This segmentation prevents unnecessary processing time while ensuring complete answers when needed.
Solution Approach 2:
The system applies partial action by performing machine reading only on passages relevant to the question type. For simple questions, it uses minimal processing to extract specific answers. For explanatory questions, it applies more extensive processing to ensure completeness. This partial/excessive action strategy optimizes the balance between processing time and answer completeness.
3Measurement precision
If neural network is used for machine reading, then answer accuracy is improved, but computational cost and processing time increase
Solution Approach 1:
The system dynamically adjusts computational resource allocation based on question type. For simple questions requiring high accuracy, it employs neural network-based extractive machine reading. For explanatory questions, it uses less computationally intensive methods to generate comprehensive answers. This dynamic resource adjustment reduces overall computational cost while maintaining high accuracy where needed.
Solution Approach 2:
The system changes computational parameters based on question characteristics. By classifying questions and adjusting the processing mode accordingly, it optimizes the balance between computational cost and accuracy. The system uses more computationally intensive methods only when necessary for simple questions, thereby reducing overall energy consumption while maintaining answer accuracy.
Data Source
AI summary
A question and answer device 100 that generates answer data in response to question data includes a question determination unit 10 that receives a predetermined question type into which a feature of the question data is classified and the passage data matching the question data, determines whether machine reading is necessary on the basis of the question type, sets the passage data as the answer data when determining that the machine reading is not necessary, and sets answer text data extracted from the passage data by the machine reading as the answer data on the basis of the question data and the passage data when determining that the machine reading is necessary.