Datasets for Aspect-Based Sentiment Analysis in Bangla and Its Baseline Evaluation

Abstract

With the extensive growth of user interactions through prominent advances of the Web, sentiment analysis has obtained more focus from an academic and a commercial point of view. Recently, sentiment analysis in the Bangla language is progressively being considered as an important task, for which previous approaches have attempted to detect the overall polarity of a Bangla document. To the best of our knowledge, there is no research on the aspect-based sentiment analysis (ABSA) of Bangla text. This can be described as being due to the lack of available datasets for ABSA. In this paper, we provide two publicly available datasets to perform the ABSA task in Bangla. One of the datasets consists of human-annotated user comments on cricket, and the other dataset consists of customer reviews of restaurants. We also describe a baseline approach for the subtask of aspect category extraction to evaluate our datasets.

Aspect Extraction from Bangla Reviews Through Stacked Auto-Encoders

“…We exploited AEs, contractive AEs, and sparse AEs, trained in the stacked fashion. All the proposed models show better precision, recall, and F1-score, with respect to the state-of-the-art works of Rahman et al [7,8].…”

“…As far as we know, the only two datasets in the Bangla language that come with aspect and polarity information are the ones published in Rahman et al [7]. As we said in the Introduction section, the authors introduced two datasets named "cricket" and "restaurant".…”

“…The need for performing accurate ABSA in the Bangla language encouraged researchers to construct datasets to analyze people's opinions in the Bangla language and to classify their sentiments, across several aspects. As far as we know, in Rahman et al [7] the authors proposed the only two datasets that can be used for benchmarking methods that address the ABSA task in the Bangla language. The authors introduced two datasets named "cricket" and "restaurant".…”

“…As far as we know, stacked AEs have never been exploited in the task of aspect extraction in the Bangla language. In particular, in [7] the authors relies on standard machine learning algorithms, such as k-NN, SVM and random forest, while in [8] the authors propose a CNN with a single convolutional layer, max pooling, and a final standard classification layer, composed by a fully connected neural network and a softmax.…”

“…The following work makes use of three AEs methods for the task of ABSA in the Bangla language to improve aspect classification performance with respect to [7,8]. As the authors did in such articles, we had to preprocess the sentences to remove some noise in the text (i.e., information not useful in performing the classification task), for instance, numbers, stop words and punctuation.…”