Vol.:(0123456789) 1 3 Journal of Ambient Intelligence and Humanized Computing https://doi.org/10.1007/s12652-020-01791-9 ORIGINAL RESEARCH Sentiment analysis of student feedback using multi‑head attention fusion model of word and context embedding for LSTM K. Sangeetha 1  · D. Prabha 2 Received: 18 November 2019 / Accepted: 18 February 2020 © Springer-Verlag GmbH Germany, part of Springer Nature 2020 Abstract Classroom teaching becomes viable and efcient based on increase in participation of the student. This can be made possible by taking needed measure by fnding the emotions of the students. Many researchers worked on emotion identifcation of students. Now-a-days sentiment analysis using deep learning models have gained good performance. Especially ensemble Long Short-Term Memory (LSTM) with attention layers gives more attention to the infuence word on the emotion. In the proposed method, input sequences of sentences are processed parallel across multi-head attention layer with fne grained embeddings (Glove and Cove) and tested with diferent dropout rates to increase the accuracy. Later in this paper, the infor- mation from both deep multi-layers is fused and fed as input to the LSTM layer. In this paper, we conclude that the fusion of multiple layers accompanied with LSTM improves the result over a common Natural Language Processing method. Keywords LSTM · Deep learning · Glove · Cove · Multi-head attention 1 Introduction The main purpose of teaching practice is to improve student learning and it has to be efective as it promotes knowledge in to students. Agarwal et al. (2011) considered that efective teaching, points have four important aspects. Especially in teachers of higher education, settings are outcomes, clarity, engagement, and enthusiasm. Poulos et al. (2008) stated that efectiveness of teaching is always depends on teachers pedagogical ability and subject knowledge. In this paper, we studied with the aim of how feedback of students becomes the efectiveness part of teach- ing learning process. Baradwaj et al. (2011) stated feedback helps the teachers to formulate better decisions on how to improve the quality of teaching. Student feedbacks are always deep and wide. They also give a comprehensive view on how their teachers encourage and educate. The student feedback gives the teacher the opportunity to feel and understand the importance in teach- ing. They also get the chance to learn about their students from the feedback settings. Also students get benefts from the system. Agarwal et al. (2011) proposed various tools to analyze and evaluate the opinion of students through feedback. Cum- mins et al. (2010) stated sentiment analysis is one of the famous and emerging technology in the feld of NLP. Senti- ment analysis (SA) evaluates the students’ opinion automati- cally by classifying them in to positive, negative and neutral class said by Vohra et al. (2013). There are several methodologies and standard tool has been developed in the evaluation of student’s feedback using machine learning techniques afrmed by Agarwal et al. (2011). There are some challenges in these tech- niques, they are (1) if the dimension of the word increases then the traditional methods failed to fnd the relationship between the words, (2) The efciency and accuracy in results depends purely based on manual feature selection, (3) pre- vious research concentrates all the words and produces the target that is a time consuming process, (4) Singlehandled mechanism is inaccurate in sentimental analysis task. In order to face the problems raised by traditional models, a proposed model depends on popular deep learning methods like embedding using Glove and Cove, attention mechanism * K. Sangeetha sangeethakalyaniraman@gmail.com D. Prabha prabha@skcet.ac.in 1 Department of CSE, Research Scholar, Anna University, Panimalar Engineering College, Chennai, Tamilnadu, India 2 Department of CSE, Sri Krishna College of Engineering and Technology, Coimbatore, Tamilnadu, India