Algorithm Inspection for Chatbot Performance Evaluation
V. Vijayaraghavan, Jack Brian Cooper, Rian Leevinson J.
Abstract
With the usage of Chatbots growing at an unprecedented rate, it is imperative that they are thoroughly tested, verified and validated. This is done to ensure that Chatbots do not fail during operation. Chatbot failures are undesirable and they often occur when the bot is provided with ambiguous or illegible input. The Chatbots have to be thoroughly tested to ensure that they do not fail under any circumstance and have mechanisms to deal with such scenarios. Although various methods are in use to test Chatbots, algorithm testing is a promising solution to the problem. This involves the use of techniques such as cross-validation, grammar and parsing, verification and validation and statistical parsing. This paper aims to explore the prominent types of chatbot testing methods with detailed emphasis on algorithm testing techniques.