Development of novel conversational agent using retrieval-based model and deep learning

Leong, Pui Huang (2021) Development of novel conversational agent using retrieval-based model and deep learning. Doctoral thesis, Universiti Teknikal Malaysia Melaka.

	Text (24 Pages) Development of novel conversational agent using retrieval-based model and deep learning.pdf - Submitted Version Download (342kB)
	Text (Full Text) Development of novel conversational agent using retrieval-based model and deep learning_unlocked (1).pdf - Submitted Version Restricted to Registered users only Download (12MB)

Abstract

The exponential development of technology have altered the mindset and the direction of knowledge retrieval. The use of the conversational agent to convey information on hehalf of humans has gained attention and has been commonly used in e-commerce, culture, medical, and education. As a result, a conversational agent was incorporated to the corporate portal, similar to engaging with the professional advisor for customer support. Despite the popularity of the conversational agents, the underlying drawbacks such as manual knowledge base creation, query misinterpretation as well as massive human intervention were still inevitable. This work automated the conventional way of performing manual knowledge base creation through the use of knowledge crawler with a one-click solution. In addition, predefined recommendations with recent popularity were introduced to minimise human error and human effort in query generation to mifigate the misinterpretation of the request. Furthermore, this work aimed to exploit the use of Deep Leaming to generate own responses to unknown queries. This research relied on the Artificial Intelligence (AI) to include Natural Tanguage Processing (NLP) by providing a paradigm focused on Refrieval-based Model and Deep Learning. The Retrieval-based Model helped the conversation agent to create better representations to match intent and content to implicitly offer quicker and more reliable answers. In contrast, Deep Learning helped the conversation agent to make smarter and better decisions to have the ability to interpret and produce its own responses. The testing was performed in the quantitative approach using Black-box Test erience (UX) Testing. The findings revealed that the conversational age: 10 select the suitable model to undergo a series of response generation which have not been seen in the conventional conversational agents. The findings from the Black-box Testing indicates the average time taken for Retrieval-based Model to generate a response was less than one second whereas the average time taken for Deep Learning to generate a response was approximately 2 seconds, In terms of accuracy, Retrieval-based Model was able to achieve high performance with 100% accuracy for Frequently Asked Questions (FAQs), 90.63% accuracy for Common Knowledge and 67.65% accuracy for Deep Learning. Apart from this, the test results from user Experience Testing indicated that 84.3% of the respondents were satisfied with the conversational agent, whereas 86% of respondents would use this conversational agent again. Both the testing showed good outcomes and proof the success of this study with enormous contributions been made in the field of artificial conversational agent. Via this innovative approach, efficiency in the growing industry could rigorously improve and indirectly capable of atracting higher pools of individuals to satisfy the service provided.

Item Type:	Thesis (Doctoral)
Uncontrolled Keywords:	Internet programming, Electronic data processing
Subjects:	Q Science > Q Science (General) Q Science > QA Mathematics
Divisions:	Library > Tesis > FTMK
Depositing User:	F Haslinda Harun
Date Deposited:	13 Jan 2023 15:39
Last Modified:	13 Jan 2023 15:39
URI:	http://eprints.utem.edu.my/id/eprint/26096
Statistic Details:	View Download Statistic

Actions (login required)

View Item