Text Coherence Analysis based on Misspelling Oblivious Word Embeddings and Deep Neural Network

Text coherence analysis is the most challenging task in Natural Language Processing (NLP) than other subfields of NLP, such as text generation, translation, or summarization. There are many methods them graph-based entity-based for short documents. However, long documents, existing perform low accuracy results which biggest challenge both English and Bengali. This because do not consider misspelled words a sentence cannot accurately assess coherence. In this paper, method has been proposed based on Misspelling Oblivious Word Embedding Model (MOEM) deep neural network. The MOEM model replaces all with correct captures interaction between different sentences by calculating their matches using word embedding. Then, network architecture used to train test model. study examines two types datasets, one Bengali English, analyze consistency sequence activities evaluate effectiveness language dataset, 7121 documents have where 5696 (80%) training 1425 (20%) testing. And 6000 1500 evaluation out 7500 efficiency compared techniques. Experimental show that significantly improves automatic detection 98.1% 89.67% Finally, comparisons models shown datasets.