Automated Essay Type Paper Marking System
No Thumbnail Available
Date
2020
Journal Title
Journal ISSN
Volume Title
Publisher
Uva Wellassa University of Sri Lanka
Abstract
Automated paper marking is a very important research tool for the education evaluation
process. Some researchers indicated that almost every study’s challenge was to get the
semantic similarity of an essay rather than keyword matching. Another major problem is
the lack of sufficient data that needed to train the system for a specific domain with a
supervised learning approach and there are some issues with the unavailability of
educator’s involvement with the scoring systems, also there were no studies that behave
like a complete system. The automated scoring or evaluation for written student responses
have been, and are still a highly interesting topic for natural language processing (NLP)
and Machine Learning (ML) research. This study is focused on building a complete
system that automates essay paper marking with a novel approach using NLP and ML.
Primarily, researchers have used a hybrid approach to get the semantic similarity between
two textual objects which contain word-vector-similarity, knowledge-based- similarity,
and word-order-similarity. As one of the main advantages, our system uses an
unsupervised learning approach, so that the system can work independently without
training for a specific subject domain. The emerging of word embedding encouraged the
calculation of the word-vector-similarity with Vector Space Model and cosine-similarity
mechanisms. On the other hand, the word-net knowledge base was used to calculate the
semantic distance between the documents and word-order-similarity played a major role
in the accuracy of the final result. Also, machine learning techniques and a vast number of
NLP techniques have been used for implementation. Besides, the proposed study contains
an OCR to identify student's handwritten characters and also a website to easily interact
with the system. In conclusion, the system was tested and evaluated with 30 samples of
essays and the manual scores given by the educators. As a result, it indicated a strong
positive correlation of (0.882) between manual scores and the system scores.
Keywords: Automated essay scoring (AES), Natural language processing (NLP), Machine
learning (ML), Optical character reader (OCR)
Description
Keywords
Computer Science, Information Science, Computing and Information Management, Education