Publication

Abstract : Visually impaired people find it difficult to study from educational websites, even though they have a lot of knowledge to provide. They are unable to feel confident about using online tools to learn because of their lack of independence. This becomes a problem for almost every student when they have to search through different websites for answers to their doubts. A chatbot is a conversational AI technology that simulates human interaction. It is more comfortable for users to interact with a chatbot if it is voice-enabled. Our proposed solution is a chatbot that is voice-enabled as well as multilingual. It can interact with more than 180 global languages, although, in this paper, we have focused on the Hindi language. It can help visually impaired people as well as normal students, as they can learn by just having a conversation with the voice bot. Our voice bot can teach students many concepts regarding data structures and algorithms. It gives the user a speech-based response received from a generated .mp3 file. For the voice-bot system, which has been trained with more than 70 intents, we have utilized the RASA NLU principle. It has been deployed onto our e-learning web application. The chatbot is compatible with all the languages mentioned in the python translate module. This chatbot recognizes and converts the speech received from the microphone into text and matches with the trained intents information, and accordingly gives the most appropriate answer and converts text-to-speech and presents the text and speaks in the language selected along with multi-intent classifications. We have made a custom pipeline and also applied various policies to it to improve pipeline performance. Nearly all of the queries are correctly answered by the voice bot, which makes communication more effective.