A Novel Mechanism for Generating FAQs in the Question-Answering System of the E-Learning Platform

• In Event: Innovative Learning Design in an Information and Communications Technology-Rich Environment: The Ongoing Story of Shanghai

Abstract

Objectives: The question-answering system is a part of e-learning platform. When students encounter difficulties, they will resort to the help section of a question-answering system. FAQs can assist the teacher to identify the difficulties for adjusting the learning process. This study presents a mechanism for generating FAQs that can help teachers to discover the difficulties of students.
Perspectives: The similarity of questions in the question-answering system plays an important role in capturing FAQ topics. The accurate measurement of similarities among questions is crucial. Any question that students input could convey information needed for the study. We also note the following two reminders: (1) If two questions contain the same terms, they convey the same or similar information. The more terms are in common, the more similar they are. (2) If two questions are semantically associated and the degree of similarity is greater than a certain threshold, then they convey the same or similar information.
Methods: An experimental approach is adopted. Based on an e-learning platform, the study groups the questions that students input. FAQs are generated through question clustering. Given the wide variety of questions, the system does not know how many topics will exist. Therefore, the clustering method does not need users to manually set the number of clusters. After a comprehensive comparison, the bottom-up hierarchical clustering method is adopted. In the process, both the word form and semantic information are considered. Furthermore, 20 students are asked for marking the result. Moreover, the teacher is asked to adjust the learning materials accordingly.
Data sources: Data were obtained from a question-answering system in an e-learning platform, from March 6, 2013 to June 6, 2013. After preprocessing the data, removing the incomplete and uncivilized ones, 518 questions were left. Therefore, 518 questions comprise the experimental data.
Preliminary findings: Experiments indicate the following preliminary findings: (1) Semantic information contributes to obtaining the meaning of questions and making the clustering results more correctively. (2) FAQs assist in identifying student difficulties in e-learning. (3) Teachers benefit from FAQs and adjust the learning materials accordingly. Good performance can then be achieved.

Author

• Lingling Meng, East China Normal University