AUTHOR=Aljohani Tahani , Cristea Alexandra I. 

TITLE=Learners Demographics Classification on MOOCs During the COVID-19: Author Profiling via Deep Learning Based on Semantic and Syntactic Representations

JOURNAL=Frontiers in Research Metrics and Analytics

VOLUME=Volume 6 - 2021

YEAR=2021

URL=https://www.frontiersin.org/journals/research-metrics-and-analytics/articles/10.3389/frma.2021.673928

DOI=10.3389/frma.2021.673928

ISSN=2504-0537

ABSTRACT=In this paper, we seek to improve User Profiling (UP), i.e. estimating the demographic characteristics of learners in MOOC platforms. We have focused on predicting the employment status of learners based on the semantic representation of text. We have compared the sequential with the popular parallel ensemble deep learning architecture for UP based on Convolutional Neural Networks and Recurrent Neural Networks, obtaining an average high accuracy of 96.3% for our best method. In following stage, we focused on predicting the gender of learners based on syntactic knowledge from the text. We have compared different tree-structured Long-Short-Term Memory models (as state-of-the-art candidates) and provided our novel version of Bi-directional composition function for existing architectures. In addition, we have evaluated 18 different combinations of word-level encoding and sentence-level encoding functions on this. Based on these results, we showed that our Bi-directional version model outperforms all other models and the highest accuracy result among our models is the one based on the combination of FeedForward Neural Network and the Stack-augmented Parser-Interpreter Neural Network (82.60% prediction accuracy). We show that such prediction models for both characteristics can achieve high accuracy, and that pre-course questionnaires to extract the UP with a high cognitive overhead for user profiling could become redundant.