Spelling Correction in Rule Based Virtual Assistant for a MSME Business

Virtual assistant is a system that can communicate like a conversation that functions as an assistant to answer user questions. Users can type the questions at the virtual assistant menu. The research conducted resulted in a system that can make corrections to user typing on a virtual assistant that manages remove, duplicate, add, and substitute spelling errors. The role of artificial intelligence is more prominent to enhance the experience of customers and ease the burden of customer service and reduce cost. Often times, users mistype or misspell a sentence to deliver messages in a quick manner. But as computers could only understand correctly typed words, Automatic Spelling Correction (ASC) methods were born. This research implements Levenshtein Distance to correct incorrect words that combines it with a rule-based virtual assistant to serve information to the users using the web. In this experiment we implemented spelling correction in a real-time virtual assistant on a MSME website, it requires a small corpus to not burden the server and the users. We try to build this spelling correction using the Symspell library on the corpus from online news article (Tempo) with alterations catered to the needs of the business. The best results were achieved through the utilization of the ”Add” error type, achieving an impressive accuracy rate of 91.67%. On average, the system successfully corrected 81.25% of one-character errors from the original input.