Statistical approach for figurative sentiment analysis on Social Networking Services: a case study on Twitter

This paper presents a system that analyzes the sentiment of figurative language contained in short texts collected from Social Networking Services (SNS). This case study sources information from tweets on Twitter and calculates the polarity of the figurative language with three different categories (i.e., sarcastic, ironic, and metaphorical tweets). As in Medhat et al. (Ain Shams Eng J 5(4):1093–1113, 2014), Nguyen and Jung (Mob Netw Appl 20(4):475–486, 2015), many related works have used a lexical-based approach (e.g., dictionary and corpus), and machine learning-based approach (e.g., decision tree, rule discovery, and probabilistic methods) to extract sentiment in a given text. This statistical approach makes use of two main features: i) Content-based, and ii) Emotion Pattern-based. We believe that this combination offers a general method to solve the current problem and easily extends for analyzing other types of figurative languages. The proposed algorithm is evaluated by using Cosine similarity to conduct an experiment over a Data set that contains about 5,000 tweets. The results show that the FIS Model (Figurative language Identification using Statistical-based Model) works well with figurative tweets with a highest achievement of 0.7813.