No abstract

The use of Bangla abusive texts has been accelerated with the progressive use of social media. Through this platform, one can spread the hatred or negativity in a viral form. Plenty of research has been done on detecting abusive text in the English language. Bangla abusive text detection has not been done to a great extent. In this experimental study, we have applied three distinct approaches to a comprehensive dataset to obtain a better outcome. In the first study, a large dataset collected from Facebook and YouTube has been utilized to detect abusive texts. After extensive pre-processing and feature extraction, a set of consciously selected supervised machine learning classifiers i.e. multinomial Naïve Bayes (MNB), multi layer perceptron (MLP), support vector machine (SVM), decision tree, random forrest, stochastic gradient descent (SGD), ridge, perceptron and k-nearest neighbors (k-NN) has been applied to determine the best result. The second experiment is conducted by constructing a balanced dataset by random under sampling the majority class and finally, a Bengali stemmer is employed on the dataset and then the final experiment is conducted. In all three experiments, SVM with the full dataset obtained the highest accuracy of 88%.

Bangla blog is increasing rapidly in the era of information, and consequently, the blog has a diverse layout and categorization. In such an aptitude, automated blog post classification is a comparatively more efficient solution in order to organize Bangla blog posts in a standard way so that users can easily find their required articles of interest. In this research, nine supervised learning models which are Support Vector Machine (SVM), multinomial naïve Bayes (MNB), multi-layer perceptron (MLP), k-nearest neighbours (k-NN), stochastic gradient descent (SGD), decision tree, perceptron, ridge classifier and random forest are utilized and compared for classification of Bangla blog post. Moreover, the performance on predicting blog posts against eight categories, three feature extraction techniques are applied, namely unigram TF-IDF (term frequency-inverse document frequency), bigram TF-IDF, and trigram TF-IDF. The majority of the classifiers show above 80% accuracy. Other performance evaluation metrics also show good results while comparing the selected classifiers.

No abstract