Polarity Detection of Online News Articles Using Deep Learning Techniques

News dissemination across various media has evolved over time. The advancement of information technology has brought news from around the globe to people’s fingertips in the form of online news articles. Technique to mine the sentiments of online news articles flooding the internet can aid in a number of financial as well as political purposes. For a long time, instrumentation of polarity detection tasks has been performed using machine learning techniques. However, these techniques are time-consuming due to high dimensionality of unstructured text. When NLP is combined with deep learning fundamentals, it shows a great improvement over the existing techniques. Hence, in this paper, we propose an approach to detect the polarity of online news articles using deep learning algorithms like CNN, RNN and other hybrid approaches by performing a comparative analysis between them. We have referred to a dataset from Kaggle, which consists of 4515 examples and contains author name, Headlines, URL of Article, Short text, Complete Article in order to do so. The news articles belong to Hindu, Indian Times and Guardian and they range from February 2017 to August 2017. The evaluation measures of different algorithms will be based on the accuracy of the result and the corresponding cost of the model In this paper, we also work on the concept of dynamic dictionaries and perform the task of polarity detection using them.