Proposed Model for Arabic Grammar Error Correction Based on Convolutional Neural Network

Deep learning and machine learning algorithms are widely used in Arabic Natural Language Processing (ANLP) aims to develop tools and techniques to process human languages in several forms like written and spoken context. The ANLP still lacks tools and applications to bridge the hole between Arabic and other languages. Furthermore, there are insufficient available resources such as dictionaries, grammatical rules, corpora, etc. Grammatical Error Correction (GEC) one of the NLP tasks seek to develop automatic tools to correct grammar and spelling, the input is incorrect words or sentences and the output become the corrected version of the same sentences. Recently, Neural Networks are used in GEC and had promising results but some improvement is still needed. The limitation of the previous studies are handcrafted and most often extracted from short sentences, also the whole previous Arabic neural approaches are used Recurrent Neural Networks (RNNs). This paper present work-in-progress for developing an Arabic GEC model based on multi convolutional layers with an attention mechanism. Moreover, proposed the incremental techniques and multi-round training model using parallel corpus to get more accuracy and fluently results, also to achieve human-level performance.