Analysis of Spelling Errors in Text Data Using Graph Neural Networks (GNNs) Based on Machine Learning

Spell checking is a step in correcting errors in a text. This research uses the Graph Neural Networks (GNNs) model to carry out machine learning-based spelling error analysis. In this research, we explored the dataset and corpus used with the aim of making it easier for Graph Neural Networks (GNNs) models to analyze spelling errors. This research uses data originating from social media and web-based text communication platforms with a total of 1000 data. The Graph Neural Networks (GNNs) model can read complex spelling error patterns based on training data and also pay attention to linguistic context and pay attention to syntax in the text so that this model can produce better spelling corrections with an accuracy rate of up to 94%. In this model there is a first, second and third test, each of which yields an accuracy. The first test produces 87% accuracy, the second test produces 90% accuracy and the third test produces 94% accuracy. After that, an evaluation was carried out using ROC (Receiver Operating Characteristic) which produced a value of 0.94. The use of machine learning in this model can help in data analysis so that data patterns are detected properly, so that spelling errors can be corrected. This research has shortcomings in terms of the complexity of the very long time in the training process due to the large amount of data and corpus used to produce good accuracy.