There is a need for a solution addressing service issues at Universitas Negeri Semarang (UNNES) helpdesk considering the limited human resources and the needs of service consumers. This work sought to apply generative-based and similarity-based answer generating models at UNNES' helpdesk in an integrated chatbot system. A fresh method in
higher education institutions, automated, context-aware responses improve response efficiency and user happiness, so contributing mostly to Using automated and contextaware answer generating, the main goal was to improve response efficiency and user satisfaction. Using the TF-IDF model for first query handling, the approach sought to rapidly access
pertinent Frequently Asked Questions (FAQ) answers. Furthermore emplo reputation was a generative model, llama RAG. This work attempts to find the best appropriate approaches for constructing such a system by means of a thorough comparison analysis of several natural language processing (NLP) techniques. This study offers three-fold contributions.
First in the framework of a FAQ-based chatbot
the study offers quantitative results on the performance of TF-IDF, BERT, Universal Sentence Encoder (USE), Sentence-BERT (SBERT), and Groove. Using criteria including Precision, Recall, F1 Score, and Accuracy, the study reveals the advantages and drawbacks of every strategy. Research prior to this indicates that for limited number FAQ datasets, SBERT shows
encouraging findings. But as the epidemic started, the support desk moved to online tools using email and live chat. Real-time question-and-answer sessions with customer support agents made possible by live chat systems enable Because of its direct and instantaneous interaction features throughout the past ten years, live chat has become a favored approach
for user involvement This function lets staff members personally handle the information or service requests for every user. Figure 2 provides further information on the live chat service duration at UNNES. The average duration of a live chat session is roughly 16 minutes and 20 seconds; the longest recorded session was 115 minutes. Figure 3 shows that customer
Service staff members respond on average five
minutes to arriving queries. This results in user impressions of the integrated service as shown in Figure 4, which reveals that out of 4004 responses about 99% of the customers rated the service as neutral, while the remaining 1% gave positive comments. These results point to the requirement of a careful review and in-depth examination of the given service
data. assistance desk averages of service duration When the amount of service requests rises but the number of customer service agents is limited, a major problem results whereby many user inquiries remain unmet or cause longer wait times, therefore possibly loweringuser happiness. Figure 3: UNNES assistance desk's first reaction times Figure assistance desk
user satisfaction level Ten percent of users of the UNNES help desk service report making repeated complaints, and thirty percent of them ask issues already covered in the Frequently Asked issues (FAQ) section. This suggests user little interaction with the FAQ material. For the UNNES service desk, this situation mostly offers two difficulties. First of all, as numerous users constantly ask the same questions, customer care staff members have more job.
Second a bigger volume of incoming
questions closely links with the time needed by customer care agents to handle all questions. The limited customer support staff members at UNNES helpdesk have made using artificial intelligence (AI) a necessary alternative to human participation for main customer service jobs. AI technology helps machines to run in a way similar to humans, therefore perhaps replacing human work For consumers looking for information, this not only lessens the
burden on customer care representatives but also increases the response time and satisfaction. An automated system guarantees users fast access to consistent and reliable information, therefore improving the general quality of services. This work aims primarily to investigate and assess several approaches for generating a score for generative answers was 0.61, suggesting strong relevance and linguistic coherence, respectively.Designed to
answer questions from staff, students, and outside users, this helpdesk is The few customer care agents have been severely strained by the growing number of questions received over the past three years (2021–2023). Dealing with this problem is crucial since good helpdesk operations are necessary to preserve high degrees of user happiness and operational effectiveness, which directly affect the reputation of the institution and the quality of services.
Conclusion
Longer wait times and a significant percentage of unanswered complaints resulting from this increasing volume emphasize the need of a scalable and efficient solution. The COVID-19 epidemic's quick shift from in-person to online services has compounded these difficulties and resulted in more user discontent and many unsolved questions. By using an artificial intelligence-driven chatbot system to improve response efficiency and customer happiness at
this paper intends to solve these problems. This work does not, however, address how the chatbot might be used in non-academic environments or how it might interact with other university services epidemic causes the helpdesk numerous particular difficulties including a restricted number of customer service staff, rising inquiry volumes, and the shift from in
Comments
Post a Comment