Improving Political Discourse Analysis on Twitter With Context Analysis

In this study, we propose a new approach to perform political discourse analysis in social media platforms based on widely used categorisation schema the field of science, namely, Comparative Manifestos Project's category schema. This has been traditionally content manifestos, giving code that indicates domain or each phrases manifestos. Therefore, work application technique Twitter, using as training data 100 publicly available annotated manifestos English with around 85,000 sentences. Furthermore, also analyse improvement 5,000 tweets could provide performance classifier already trained Finally, have analysed 2016 United States presidential elections Twitter proposed approach. As our main finding, able conclude both datasets (political and tweets) can be combined order achieve better results, achieving improvements F-Measure more than 15 points. Moreover, if contextual information such previous tweet affiliation transmitter improve classifier's it proven for classification, introducing novel method parties representation finding adding leaning does its performance.