Chatbot Implementation in Customer Service Industry through Deep Neural Networks
Himanta Dihingia, Sakil Ahmed, Dulusmita Borah, Suraj Gupta, Kuldeep Phukan, Monoj Kumar Muchahari · 2021 International Conference on Computational Performance Evaluation (ComPE) · 2021
It is the age of artificial intelligence (AI). As AI, machine learning, and Deep Learning (DL) have been progressing, machines have also advanced to emulate humans. A good example of such a machine is chatbots, which can talk like human, uses AI and NLP, can be a virtual assistant for customer experience. This paper provides a implementation of how chatbots work and the implementation of its techniques that can be used to create them. It compares and contrasts the current chatbots, as well as their strengths and weaknesses. Comparison is based on the functionalities and technical requirements of the most common chatbot application systems. According to research, nearly 75% of consumers have had bad customer service and coming up with substantive, lengthy, and insightful answers remains a difficult task. We used an encoder-decoder attention mechanism design to build the Seq2Seq AI Chatbot. This encoder-decoder makes use of LSTM cells in Deep RNNs. Instead of a specific chatbot that is used for a certain purpose, we wanted to design a general chatbot that can conduct general discussions with us like a friend. However, the model we'll employ can be trained on other datasets for different purposes. We have built a chatbot to talk about everyday conversations.