Adaptive Question Generator Using Perturbation Resistance for Skill Assessment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional learning support systems face challenges in providing questions tailored to individual learner skills, as they rely on vocabulary levels and answering times, which can't distinguish between confident and random correct answers, and require significant time and labor to create, limiting learning efficiency.
Innovation Solution
A learning support system that uses a computer with a processor and memory, featuring a question database, learning history database, question generator, and achievement degree estimator, which adjusts question difficulty by perturbing words based on learner performance, generating questions suited to individual skill levels and improving learning efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If vocabulary level is used to determine question difficulty, then questions can be categorized by difficulty level, but it cannot distinguish between learners with same vocabulary knowledge but different actual language ability
Solution Approach 1:
The patent changes the parameter for determining question difficulty from static vocabulary level to dynamic perturbation resistance. Instead of relying on pre-defined vocabulary levels, the system applies various types of perturbations (synonym replacement, antonym replacement, modification) to words and measures how well learners can identify the perturbed word, thereby adapting to individual skill levels more accurately.
2Productivity
If answering time is used to estimate ability, then quick answers may indicate high ability, but it cannot distinguish between confident correct answers and random guessing
Solution Approach 1:
The patent introduces feedback mechanisms where learners receive information about their perturbation identification performance. The system tracks which perturbation types a learner struggles with and uses this feedback to select appropriate perturbation types for subsequent questions, thereby improving both measurement precision and learning efficiency through adaptive feedback loops.
3Ease of manufacture
If artificial educational materials are created for questions, then questions can be tailored to specific learning objectives, but it requires enormous time and labor to create and maintain the question database
Solution Approach 1:
The patent uses copying by applying perturbation transformations to existing words in educational materials rather than creating entirely new questions. Instead of manually crafting each question, the system copies existing content and applies systematic perturbations (synonym/antonym replacement, modifications) to generate new questions automatically, greatly reducing creation effort while maintaining adaptability.
4Adaptability or versatility
If raw English materials from magazines and news are used, then diverse and authentic learning content can be provided, but it cannot be easily adapted to individual learner skill levels
Solution Approach 1:
The patent introduces dynamics by making the perturbation application adaptive based on learner performance. The system dynamically adjusts which perturbation types are applied to words in raw English materials based on real-time assessment of learner ability, allowing authentic content to be automatically adapted to individual skill levels without manual intervention.
Data Source
AI summary
In order to provide questions suited to skill levels of a learner, it is provided a learning support system which consists of a computer including a processor and a memory, comprising: a question database; a question generator; an answer acquisition module; a scoring module and an achievement degree estimator. The achievement degree estimator compares a number of types of perturbation within the questions to which a correct answer has been given with a predetermined second threshold value. The question generator generates a word changed in accordance with the perturbation of a different type in a case where the number of types of perturbation within the questions to which the correct answer has been given is smaller than the predetermined second threshold value, and sets a question relating to the generated word as a candidate for the question.