A Fuzzy domain ontology combined with Support Vector Machine (SVM) was applied to automate the online review classification in the work and achieved an accuracy of 82.7%. The authors obtained 98.22% accuracy for combined reviews, and 95.345% and 96.145% accuracy for the positive and negative reviews, respectively. A Convolutional Neural Network (CNN) based model for feature-based opinion mining from customer reviews in the hotel domain was developed in. We conclude this paper finally in Section 5.įor mining public opinion, especially from hotel reviews, a significant amount of research has been performed over the years. The results and discussion are explained in Section 4. In Section 3, the materials and methodology of this paper are described. In Section 2, a brief overview of the related works in the hotel reviews classification domain is presented. The rest of the paper is organized as follows. And finally, we analyzed the performance of several Deep Neural Network (DNN) based models, such as LSTM, BiLSTM, GRU, BiGRU, and a hybrid architecture of CNN-LSTM with our proposed Attention-LSTM model for mining public opinion in the hotel reviews domain. Secondly, we proposed an Attention-based Long Short Term Memory (Attention-LSTM) network for categorizing positive and negative opinions. To accomplish our objective, first, we developed word vectors using the Word2Vec model from an existing hotel reviews dataset, and then applied a transfer learning technique to develop word vectors for our gathered hotel reviews dataset. The key contribution of this paper is to mine public opinion from the hotel reviews domain. In this paper, we implemented the above technique to build an effective corpus for hotel reviews classification. Because once the large corpus is developed then it can be reused to train other corpora which are comparatively small in size in less computational time. On the other hand, Deep Learning (DL) techniques have gained immense popularity because of the lower feature engineering and expressive power of computations in NLP tasks than traditional models.įor effectively mining public opinion, especially from a domain like hotel reviews, creating a large corpus from a huge number of reviews and using that corpus to build a new corpus consisting of a small number of reviews can reduce the computational time and improve the accuracy. However, data sparsity is a concerning issue for these models.
Support Vector Classifier (SVC) technique was used by the author in to classify the textual data accurately. Logistic Regression (LR) and Naive Bayes (NB) Machine Learning (ML) approaches were applied in, for textual data analysis. The research was conducted by applying Support Vector Machine (SVM) using TF-IDF features and Bag of Words (BOW).
For instance, a supervised machine learning method was proposed for classifying hotel reviews in the work. To do that, previous researchers applied a variety of Machine Learning (ML) and Deep Learning (DL) based techniques for classifying online hotel reviews. So, developing an efficient method for processing a large number of online reviews would be quite beneficial. Performing such an extensive study will certainly cost the consumers precious time. As a result, a huge number of reviews are explored by consumers who devote their adequate mental energy to reaching a specific opinion. Several reviews contain biased information or are simply pointless, while on the contrary, other reviews are very helpful in objective evaluation. Furthermore, the consistency of the quality of reviews is another important issue. “Previous studies also confirmed the impact of online hotel reviews on consumers and the hotel industry as well”. There is a report which indicates that 95% of customers before making their online hotel bookings browse online hotel reviews. However, these online reviews help consumers to shape their travel experiences and represent electronic word-of-mouth (eWoM). For instance, various travel sites like () and () contain a huge number of travel reviews, scores, ratings, and feedback. Mining public opinions can be a tricky problem as there are a vast number of reviews available online.
0 kommentar(er)
