A Convolution-LSTM-Based Deep Neural Network for Cross-Domain MOOC Forum Post Classification

Abstract

Learners in a massive open online course often express feelings, exchange ideas and seek help by posting questions in discussion forums. Due to the very high learner-to-instructor ratios, it is unrealistic to expect instructors to adequately track the forums, find all of the issues that need resolution and understand their urgency and sentiment. In this paper, considering the biases among different courses, we propose a transfer learning framework based on a convolutional neural network and a long short-term memory model, called ConvL, to automatically identify whether a post expresses confusion, determine the urgency and classify the polarity of the sentiment. First, we learn the feature representation for each word by considering the local contextual feature via the convolution operation. Second, we learn the post representation from the features extracted through the convolution operation via the LSTM model, which considers the long-term temporal semantic relationships of features. Third, we investigate the possibility of transferring parameters from a model trained on one course to another course and the subsequent fine-tuning. Experiments on three real-world MOOC courses confirm the effectiveness of our framework. This work suggests that our model can potentially significantly increase the effectiveness of monitoring MOOC forums in real time.

Temporal analysis of multimodal data to predict collaborative learning outcomes

“…When using LSTMs with outcome data, they can be used to find successful and unsuccessful strategies (Akram et al, 2018), predict the quality of a presentation (Li, Wong, & Kankanhalli, 2016) or predict learning outcomes (Okubo, Yamashita, Shimada, & Ogata, 2017). LSTMs can more accurately predict learning than common methods like Bayesian knowledge tracing (Piech et al, 2015) On the other hand, they have also been used to provide interventions during the learning process by providing learning recommendations (Zhou, Huang, Hu, Zhu, & Tang, 2018), predicting engagement and affect (Botelho, Baker, & Heffernan, 2017;Le et al, 2018), collaboration actions (Shibata, Ando, & Inaba, 2017), learning state (Wang, Sy, Liu, & Piech, 2017) and forum post relationships (Wei, Lin, Yang, & Yu, 2017). The longer history afforded through the use of LSTMs provides a way for the predictions to consider temporal patterns.…”

Co-Training Semi-Supervised Deep Learning for Sentiment Classification of MOOC Forum Posts

“…Ramón et al utilized a CNN to detect the positive or negative polarity of learners' opinions regarding the exercises they solved in an intelligent learning environment [6]. Wei et al proposed a transfer learning framework based on CNN and LSTM to automatically identify the sentiment polarity of MOOC forum posts [13].…”

“…Moreover, CNN can explicitly capture the local contextual information between words, sub-words, and characters, which can be regarded as the deep feature extraction of the text. Quite a lot studies have proven that CNN is powerful for MOOC sentiment classification [6,13]. Thus, we combine each view and a CNN.…”

“…In many real-world applications, the amount of labeled data is very small compared to that of unlabeled data. Moreover, labeling a large amount of data is expensive and time consuming [13].…”

Artificial Intelligence to Enhance Learning Design in UOW Online, a Unified Approach to Fully Online Learning

“…Often, this is referred to as resembling the way the human brain processes input (e.g., a picture) to produce an output (e.g., recognition of a cat). In automatic text classification, such as semantic tagging, labelling or determining a sentiment, deep learning models have proven to be very effective [36,37].…”