Auto-tagging of Short Conversational Sentences using Natural Language Processing Methods

In this study, we aim to find a method to auto-tag sentences specific to a\ndomain. Our training data comprises short conversational sentences extracted\nfrom chat conversations between company's customer representatives and web site\nvisitors. We manually tagged approximately 14 thousand visitor inputs into ten\nbasic categories, which will later be used in a transformer-based language\nmodel with attention mechanisms for the ultimate goal of developing a chatbot\napplication that can produce meaningful dialogue. We considered three different\nstate-of-the-art models and reported their auto-tagging capabilities. We\nachieved the best performance with the bidirectional encoder representation\nfrom transformers (BERT) model. Implementation of the models used in these\nexperiments can be cloned from our GitHub repository and tested for similar\nauto-tagging problems without much effort.\n