Article Preview
Top1. Introduction
In recent years, with the rapid evolution of artificial intelligence and the continuous progress in natural language processing (NLP) technology, intelligent customer service dialogue systems have emerged as a focal point in the NLP field (Nie et al., 2021). Originating from expert knowledge bases, these systems are designed to assist users in answering questions and enhance response efficiency (Ji & Zhang, 2023). Considered a valuable product for human-machine interaction, intelligent customer service dialogue systems have garnered substantial attention from both industry and academia. An intelligent customer service dialogue system is a computer program designed to parse and comprehend user queries through modules such as natural language understanding (NLU), natural language generation, and dialogue management. Importantly, this entire process involves minimal human intervention (Wang et al., 2023a or b). In comparison to traditional human-operated customer service, intelligent customer service dialogue systems exhibit robust capabilities, enabling them to access high-quality, accurate, comprehensive, in-depth, and real-time information from extensive knowledge databases effectively. This capability bridges the significant gap between users' precise and diverse information needs and the vast scale of internet data (Cai et al., 2016). Additionally, they enhance user interaction, ultimately improving the overall user experience. Intelligent customer service holds significant application potential, making research on enhancing the intelligence of customer service systems a focal point in current studies (Xiao & Kumar, 2021).
The earliest research on intelligent dialogue systems can be traced back to 1950 when Alan Turing, known as the “father of artificial intelligence,” posed the question of whether machines could communicate in natural language like humans. This question later evolved into the famous Turing Test, considered the ultimate goal of artificial intelligence (Chen et al., 2017). The first chatbot system, developed by Joseph Weizenbaum at the Massachusetts Institute of Technology in 1966, was ELIZA (Al-Rfou et al., 2016), designed to emulate a psychotherapist for clinical therapy. This dialogue system primarily relied on keyword matching and manually crafted response rules. Subsequently, in 1988, a chat dialogue system named UNIX Consultant was developed by researchers, including Robert Wilensky from the University of California, Berkeley. It aimed to assist users in learning UNIX system administration. In 1995, Richard S. Wallace, inspired by ELIZA, created the renowned ALICE system, known for its heuristic template matching dialogue strategy. It is considered one of the best-performing systems of its kind. The ALICE system rapidly constructs a robust and relatively flexible retrieval-based dialogue system using the artificial intelligence markup language (AIML). The methods proposed in the aforementioned literature are all based on matching mechanisms to achieve intelligent dialogue. Even in cases where question-answer pairs have not been preconfigured, this type of dialogue system still requires a significant amount of manual effort (Lowe et al., 2015). Currently, the challenge in achieving highly intelligent dialogue systems lies in the fact that customer conversations typically involve context and subtle language nuances, making it difficult for existing intelligent customer service systems to effectively simulate human thought processes during conversations. Furthermore, dialogue content may consist of incomplete sentences, topic shifts, and disorganized discourse, further complicating the ability of current intelligent customer service systems to fully meet user requirements (Bai et al., 2022). Therefore, further research and optimization of intelligent customer service systems are considered imperative.