Grading students' answer sheets using Microsoft Azure's Computer Vision (OCR), and Natural Language Processing services (KeyPhrase Extraction and Sentiment Analysis)
A grading system that uses Artificial Intelligence to identify scanned student answer booklets, categorize the identified text, and give a grade based on a provided scoring guide.
- Upload scanned answer booklets and marking guide and AI grades based on provided scoring
- View result of graded tests
- JavaScript
- React
- MUI
- Styled-components
- Project Management: https://trello.com/b/kYooIU88/ai-grader
- Users input scanned/photographed documents (answer booklets) into the system.
- Azure's OCR Read API identifies text
- Key phrase etraction is used to extract key phrases.
- The system matches key phrases from answer booklets to key phrases from marking guides and scores based on that comparison.
- How do we add custom key phrases so we can extract key phrases that may not be picked up automatically by the azure service.
- Would the marking guide be provided as a set of key points/phrases to be used by the system's "comparison mechanism" or would it be put into the system for the OCR to identify key phrases contained therein?
- How would grading be done? Is it that when matching key phrases are identified, they would just be a given a "correct" or "wrong' score?
- There would need to be a way to determine level of correctness so that students that only got part of a question correct can be given some marks out of the total achievable marks(e.g. 3/5) instead of a generic "correct" or "wrong" score.
- The system may later be developed to monitor performance in certain courses and display results in insightul format for various analytical uses.
- The system may later be able to search for answers to questions on the internet and hence provide more accurate grading.
System can currently only grade simple written text. Cannot grade diagrams or math Some Identified possible hinderances to the project include:
- Ability of OCR to identify illeligible handwriting
- Accuracy of system in correctly matching answer sheet key phrases to correct (marking guide) key phrases
- https://docs.microsoft.com/en-gb/learn/modules/read-text-computer-vision/
- https://docs.microsoft.com/en-us/azure/cognitive-services/text-analytics/how-tos/text-analytics-how-to-keyword-extraction
- https://docs.microsoft.com/en-us/azure/search/cognitive-search-skill-keyphrases
- https://docs.microsoft.com/en-us/azure/cognitive-services/computer-vision/vision-api-how-to-topics/call-read-api
- https://docs.microsoft.com/en-us/azure/cognitive-services/text-analytics/quickstarts/client-libraries-rest-api?tabs=version-3-1&pivots=programming-language-javascript#key-phrase-extraction
- https://www.npmjs.com/package/@azure/ai-text-analytics
- https://github.com/Azure-Samples/cognitive-services-quickstart-code/blob/master/javascript/ComputerVision/ComputerVisionQuickstart.js