The impact of natural language preprocessing on big data sentiment analysis

The sentiment analysis determines peoples’ opinions, sentiments and emotions by classifying their written text into positive or negative polarity. The sentiment analysis is important for many critical applications such as decision making and products evaluation. Social networks are one of the main sources of sentiment analysis. However, the huge volume of data produced by social networks requires efficient and scalable analysis techniques to be applied. The MapReduce proved its efficiency and scalability in handling big data, thus attracted many researchers to use the MapReduce as a processing framework. In this paper, a sentiment analysis method for big data is studied. The method uses the Naïve Bayes algorithm for classifying texts into positive and negative polarity. Several linguistic and Natural Language Processing (NLP)preprocessing techniques are applied on a Twitter data set, to study their impact on the accuracy of big data classification. The preformed experiments indicates that the accuracy of the sentiment analysis is enhanced by 5%, yielding an accuracy of 73% on the Stanford Sentiment data set.