A Chatbot based on Deep Neural Network and Public Cloud Services with TJBot Interface
Michal Povinský, Miroslav Melicherčík, Vladimír Siládi · 2019
This paper describes an adaptation of the IBM TJBot as a chatbot interface. TJBot is an open source project designed to use artificial intelligence services in user friendly way. Primary it has been developed to be used with IBM Watson services. The adaptation has been done in three steps. In the first step deep neural network (DNN) based chat has been designed. Subsequently, three DNN where designed to perform experiments. They differed in training sets. The second step presents a join of DNN based chat with IBM Watson services (Speech-To-Text, Text-To-Speach). Originally, IBM Watson provides these services for limited number of languages. Slovak language is not included, too. Google cloud services fill this gap quite in good manner. This led to replacement of the IBM Watson services by the Google services. Finally, the chatbot is able to communicate in multiple languages, Slovak language including. Any non-English conversation has to be translated to English language and vice-versa by the Google translate service. This modified chatbot was tested by chat with randomly selected users.