Question Generation System with Answer Presence Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing question generation technologies often produce question sentences that include answers already present in the original text, lacking mechanisms to differentiate between questions with and without included answers.
Innovation Solution
An information processing apparatus utilizing a machine learning model to generate question sentences and classify them as either questions or extracted text based on the presence of answers in the analysis target sentence, ensuring that questions are only generated when answers are not included.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a question generation technology is used to generate question sentences from original text, then question sentences can be generated for use as scenarios and learning data, but question sentences in which an answer has already been described in the text may be generated
Solution Approach 1:
The system segments the question generation task into two distinct processes: (1) generating candidate question sentences from the original text, and (2) classifying each candidate to determine whether its answer is contained in the text. This segmentation allows the system to maintain high question generation productivity while adding a separate classification mechanism to ensure reliability.
Solution Approach 2:
A classification model serves as an intermediary between the question generation component and the final output. This intermediary classifies each generated question as either 'answer-in-text' or 'answer-not-in-text', enabling the system to filter appropriate questions while maintaining the original generation capability.
2Productivity
If existing question generation technology is applied, then question sentences are generated efficiently, but there is no mechanism to differentiate between questions with and without included answers
Solution Approach 1:
The classification model performs multiple functions: it differentiates between answer-present and answer-absent questions, enables selective filtering of generated questions, and provides metadata for downstream applications. This multi-functionality adds capability without requiring separate specialized systems for each function.
Solution Approach 2:
The system uses the classification results to automatically determine which generated questions should be outputted. The classification model serves the question generation process by providing automatic filtering and categorization, eliminating the need for manual review or additional complex verification mechanisms.
Data Source
AI summary
An information processing apparatus includes processing circuitry configured to receive an analysis target sentence as an input and generate a generation sentence, the generation sentence being a sentence regarding content of the analysis target sentence, and a classification type, the classification type being information indicating whether or not the generation sentence is a question sentence for information in which an answer is not included in the analysis target sentence, using a machine learning model learned in advance, and output the generation sentence and the classification type generated.