Automated Dialogue-Based Response and Resolution of Conversational IT Tickets Using Deep Neural Networks

In this article, we have proposed an automated dialogue-based resolution of IT tickets. The main challenge for this study is that no direct conversational IT ticketing dataset is available. We have used the AirDialogue dataset, which contains conversations between customers and agents on flight bookings. The dialogues related to IT-related booking issues are only considered as they resemble IT tickets. The dialogue dataset is also unsupervised. Therefore we have classified the dataset using Sentiment analysis and Topic modeling techniques. After the preprocessing and classification, we employed Deep neural networks such as DNN, LSTM, and BERT to generate the responses automatically. Our results show that BERT has outperformed the other Neural Networks with an accuracy of 99.79%.