Indonesian generative chatbot model for student services using GPT

Abstract

The accessibility of academic information greatly impacts the satisfaction and loyalty  of university  students.  However,  limited  university  resources  often hinder students from conveniently accessing information services. To address this  challenge,  this  research  proposes  the  digitization  of  the  question-answering  process  between  students  and  student  service  staff  through  the implementation  of  generative  chatbot.  A  generative  chatbot  can  provide students   with   human-like   responses   to   academic   inquiries   at   their convenience.  This  research  developed  generative  chatbot  using  pre-trained GPT-2   architecture   in   three   different   sizes,   specifically   designed   for addressing  practicum-related  questions  in  a  private  university  in  Indonesia. The  experiment  utilized  1288  question-answer  pairs  in  Indonesian  and demonstrated the best performance  with a BLEU score of 0.753, signifying good performance accuracy in generating text despite dataset limitations.