Objective : To automate assessment of short free text answers
Concept : Developing ML model trained based on human-labelled answer evaluation dataset of short free text answers to provide automated assessment
Issues addressed: Assessment plays a central role in any educational process, because it is a common way to evaluate the students' knowledge regarding the concepts related to learning objectives. But the expense and logistics of scoring them reliably often present a barrier to their use. Automating the process of evaluating these assessments can help reduce the burden on the evaluators.
Project Plan:
| Step | Know-how | Status |
|---|---|---|
| Collection of data on short free text answers to questions using the student assessment app | Done | |
| Develop web-portals to evaluate the short text answers to be used for training and evaluating the ML model | Web Development | Done |
| Implement short text assessment using semantic similarity bwtween the free text answers and ground truth as input | Keyword analysis, natural-language processing and Information mining techniques | In-progress |
| Conceptualize and integrate the designed model with an Android app (Deployment) | Android Development | Not Started |
Details: Automatic evaluation is preferred to manual assessment to avoid monotonic, bias errors and conserves time for other activities. Hence automatic assessment is vital for educational system. Keyword analysis, natural-language processing and Information mining techniques are the main approaches adopted for text assessment.
Next steps:
- Implementing short text assessment using semantic similarity between the ground truth and answer by student
- Develop an unsupervised method for short text assessment using clustering based approaches