Student feedback analysis is time-consuming and laborious work if it is handled manually. This study explores the use of a new deep learning-based method to design a more accurate automated system for analysing students’ feedback (called DTLP: deep learning and teaching process). The DTLP employs convolutional neural networks (CNNs), bidirectional LSTM (BiLSTM), and attention mechanism.To the best of our knowledge, a deep learning-based method using a unified feature set, which is representative of word embedding, sentiment knowledge, sentiment shifter rules, linguistic and statistical knowledge, has not been thoroughly studied with regard to sentiment analysis of student feedback. Furthermore, DTLP uses multiple strategies to overcome the following drawbacks: contextual polarity; sentence types; words with similar semantic context but opposite sentiment polarity; word coverage limit of an individual lexicon; and word sense variations. To evaluate the DTLP, we conducted an experiment on a large volume of students’ feedback. The results showed (i) DTLP outperforms the existing systems in the field, (ii) DTLP that learns from this unified feature set can acquire significantly higher performance than one that learns from a feature subset, (iii) the ensemble of sentiment shifter rules, word embedding, statistical, linguistic, and sentiment knowledge allows DTLP to obtain significant performance, and (iv) an attention mechanism into CNN-BiLSTM improves the performance of DTLP. In addition, the deployed method looks for potential causes behind student feedback.
After some time of lockdown experiences, limited attention for feedback and the absence of feedback digitalization frameworks suggests rethinking traditional feedback practices toward post-pandemics digital/hybrid education. This research surveyed feedback digitalization needs in the context of online education in high education institutions in the Netherlands and Germany during the COVID-19 pandemic. The dimensions surveyed included preferences for feedback such as typology of feedback (e.g., cognitive, behavioral, etc.), formats (e.g., written, audio, video), online instruments, and features for communicating feedback. The results suggest that online instruments supporting features for effortless interactivity are among the highly preferred digital options for giving/receiving feedback. When given online, inclusive formats of feedback that inform learners not only about their own but also peer performance were also found to be among highly rated options. The increased need for inclusive feedback with peers, however, may also negatively affect students’ mastery orientations. Thus, balancing online with offline approaches should also be recommended when considering feedback digitalization approaches.
Purpose: This research aims to explore digital feedback needs/preferences in online education during lockdown and the implications for post-pandemic education.Design/methodology/approach: An empirical study approach was used to explore feedback needs and experiences from educational institutions in the Netherlands and Germany (N = 247) using a survey method.Findings: The results showed that instruments supporting features for effortless interactivity are among the highly preferred options for giving/receiving feedback in online/hybrid classrooms, which are in addition also opted for post-pandemic education. The analysis also showed that, when communicating feedback digitally, more inclusive formats are preferred, e.g. informing learners about how they perform compared to peers. The increased need for comparative performance-oriented feedback, however, may affect students' goal orientations. In general, the results of this study suggest that while interactivity features of online instruments are key to ensuring social presence when using digital forms of feedback, balancing online with offline approaches should be recommended.Originality/value: This research contributes to the gap in the scientific literature on feedback digitalization. Most of the existing research are in the domain of automated feedback generated by various learning environments, while literature on digital feedback in online classrooms, e.g. empirical studies on preferences for typology, formats and communication channels for digital feedback, to the best of the authors’ knowledge is largely lacking. The findings and recommendations of this study extend their relevance to post-pandemic education for which hybrid classroom is opted among the highly preferred formats by survey respondents.