Intent Pattern Discovery for Academic Chatbot - A Comparison between N-gram model and Frequent Pattern-Growth method
Suraya Alias, Mohd Shamrie Sainin, Tan Soo Fun, Norhayati Daut · 2019 IEEE 6th International Conference on Engineering Technologies and Applied Sciences (ICETAS) · 2019
We present our findings in user's Intent Pattern Discovery for designing a conversational agent (chatbot) in the academic-related domain. We developed our chatbot corpus using social messaging dataset of the conversation between students and the internship coordinator to train our Academic Chatbot model. We compare our unsupervised approach FIP or Frequent Intent Pattern against the N-gram model to investigate the efficiency of our model to extract significant and descriptive user's intent pattern with limited vector size. Experiment results show promising intent rules discovery with 0.9 confidence value with term vector reduction size of 78% against the Bigram model. This finding has given the insight to discover the essence from a social conversation for our Academic Chatbot in order to understand a user's intention in a social conversation.