Natural language processing using text augmentation for chatbot

• Published in: 2022 International Conference on Artificial Intelligence and Data Engineering (AIDE) pp. 248 - 251
• Main Authors: J, Praveen Gujjar, H R, Prasanna Kumar
• Format: Conference Proceeding
• Language: English
• Published: IEEE 22.12.2022
• Subjects: Adversarial attacks; Artificial neural networks; Chatbot; Chatbots; Data engineering; Data models; Libraries; neural network; NLPAug; Semantics; text augmentation; Tokenization
• DOI: 10.1109/AIDE57180.2022.10060406

Chatbot is an application that helps in virtual interactions. Chatbot simulates the interaction in a more effective way by natural language processing. An artificial neural network makes better understanding of the user intent and entity. There are two broad classifications for chatbot i.e rule based and conversational AI chatbot. In the case of rule based chatbot, for all user inquiries automated responses are generated based on set of rules. Whereas in conversational AI chatbot, it requires better natural language processing for understanding user inquiries. Text augmentation provides the platform for the chatbot system to train more on the limited amount of the dataset. This paper focuses more on the text augmentation techniques for creating the robust chatbot system using the text augmentation open source library NLPAug. The recent studies outline that machine learning models are more vulnerable to even for small adversarial attack. The result shows that NLPAug perform the text augmentation in three levels those are character level, word level and sentence level to avoid the adversarial attacks. Text augmentation enables the chatbot to provide accurate result for the user query. Multimedia files can be augmented in the future work.