Bangla Sentence Correction Using Deep Neural Network Based Sequence to Sequence Learning
View/ Open
Date
2018-09Author
Islam, Sadidul
Sarkar, Mst. Farhana
Hussain, Towhid
Hasan, Md. Mehedi
Metadata
Show full item recordAbstract
Development in deep neural networks in particular to natural language processing has motivated researchers to apply these techniques in solving challenging problems like machine translation, automatic grammar checking, etc. In this paper, we address the problem of Bangla sentence correction and auto completion using decoder-encoder based sequence-to-sequence recurrent neural network with long short term memory cells. For this purpose, we have constructed a standard benchmark dataset incorporating mis-arrangement of words, missing words and sentence completion tasks. Based on the dataset we have trained our model and achieved 79% accuracy on the test dataset. We have made all our methods and datasets
available for future use of the other researchers from: https://github.com/mrscp/bangla-sentence-correction. An online tool have also been developed based on our methods and readily available to use from: http://brl.uiu.ac.bd/s2s.
Collections
- B.Sc Thesis/Project [82]