Intelligent and Proactive Approach for The Optimal Handling of Low Chatbot Quality of Services (CQoS)
- Publication Type:
- Thesis
- Issue Date:
- 2022
Open Access
Copyright Clearance Process
- Recently Added
- In Progress
- Open Access
This item is open access.
Recently, the chatbot has evolved into a trending topic in the area of computer science. The rapid growth of intelligent chatbots as conversational agents with artificial intelligence has recently attracted much research attention. This significant increase in the use of chatbots across different domains, such as education, business, and health care, raises a problematic issue, this being the quality of the responses provided by the chatbot. Although most of the research studies attempted to build a chatbot that provides an intelligent response, in some cases, a chatbot might not understand the end-user’s request, which leads to producing inappropriate utterances that cause a negative user experience and conversation breakdown. While several studies focus on dialogue breakdown detection, they still face several challenges, such as the lack and bias of human annotation for the dataset. Also, when they detect a dialogue breakdown point, they do not provide a solution to handle the breakdown. In the current literature, there is no model to determine the quality of responses from a chatbot to make intelligent and proactive decisions to transfer the conversation from the chatbot to a live agent. To tackle these challenges, in this thesis, we developed intelligent, automated, and data-driven approaches to address the aforementioned research issue of determining the chatbot quality of service (CQoS) and make proactive and intelligent decisions as to when to transfer the control of the conversation to a live agent. Various machine learning approaches are proposed to detect CQoS, including supervised and unsupervised approaches. Also another key aspect is considered, which is the human thinking and reasoning using the fuzzy logic detection model. Importantly, the use of a sentiment score is introduced to trigger the breakdown without the need for annotated dataset. The proposed solutions are evaluated using realtime datasets. The key finding of our research was based on the evaluation process. We concluded that our proposed method for modeling CQoS outperforms other similar methods. Also, based on the evaluation process, the deep learning model was able to more accurately detect the need for handover mechanism compared with the other models.
Please use this identifier to cite or link to this item: