Nepali Spelling Checker

Abstract:

Communication is a great deal in today’s era. Be it composing emails, be it writing essays for higher education, be it maintaining relationships with your loved ones through text, be it searching for problems on the internet, we can’t deny the fact that there’s a chance of spelling mistakes. Misspelled words can provide wrong information to readers and may create a negative impact on them. So, a spell checker plays a significant role in this area. Spell Checker detects and suggests the misspelled words. Spelling errors run across all the languages, not only in English. Detecting and identifying misspelled words can be even more tedious in the context of the Nepali Language since not much research has been done in the past in comparison to the English Language. In this project, we have proposed a spell checker that detects and corrects the misspelled words in the Nepali Language, especially in the health domain. Nepali Spell Checker is a sequence-to-sequence model based on GRU which solves the vanishing gradient problem. GRU takes care of the information from the past without forgetting it through time and eliminates information that is irrelevant to the prediction. Our model is based on GRU which is trained for 10 epochs and gives an accuracy of 75.11%. Keywords: Gated Recurrent Unit (GRU); Levenshtein Distance; Natural Language Processing (NLP); Nepali Spell Checker