Italian Political Discourse Analysis. Opinion and Feeling

1. INTRODUCTIONThe goal of an interdisciplinary approach as political discourse analysis, with natural language processing tools (NLP), is to define and explain various discourse contexts, in this case, Silvio Berlusconi's resignation of (November 11th, 2011) and Mario Monti's appointment as Prime Minister of the Italian Government (November 16th, 2011) - two events reflected in the written press, during October-December 2011. Most studies in this regard have focused mainly on three analytical aspects: first one focusing on the emotional dimension identified by how people use language (the discourse means people use to express their feelings); the second one, centering upon understanding the relationship between speech in a certain area and outside it and the third - the role of linguistic structure (here, the Italian language) as a means of communication. Linguists most often treat language as an abstract object that can be measured without reference to social and political concerns, whatever may their nature be.Content analysis based on the theories of Osgood et al. - but also Taboada et al. - calls for a very laborious methodology in obtaining objective results that can be validated by other scientific methods. In this paper, we discuss communication paradigms, especially those that support linguistic discourse interpretation based on natural language processing by means of the computational system GETARUNS (General Text And Reference Understanding System), which can complete access to extra-linguistic knowledge. We are considering three critical aspects for a successful evaluation of the political text from the written press: the creation of large amounts of data to train the proposed system as a method of analysis, obtaining important results in terms of semiotics, semantics and lexicology, syntax and pragmatics, and improving the so far implemented heuristics.A set of criteria for texts selection complete our assumption, i.e. that this corpus (Italian political articles from the written press) will be sufficient for extracting opinion and feeling1 with GETARUNS system. After the texts processing, at the end of the analysis, we get some data output, viable at the sentence level. In particular, this analytical approach can be checked by using tools such as BOWs (Bag of Words Approaches) to the end of recovering information - an aspect still insufficiently dealt with. Such approaches, like BOWs, are based on the matching of several keywords, the existence of certain ontologies, and on the search of concepts by means of SentiWordNet (Sentiment Analysis and Opinion Mining with WordNet), starting from a text. The system uses the input words, analyzing them syntactically and semantically. Any search based on keywords and Bows is vitiated by the failure in understanding precisely the processed word, set next to other words - fact that many times changes its meaning2: negation at different syntactic levels; verb or adverb negation; the conditional mood; the double negation with copulative verbs, the use of modal verbs etc.In order to deal with these linguistic elements, we propose a sentence level analysis, starting from a tense political context, with syntactic elements in a tree-like pattern representation. Within the GET ARUNS system there have been implemented various computational functions, necessary for the semantic evaluation called RTW, but also for other semantic tasks3. The system provides as output texts .xml format, where each sentence is a list of attribute-value pairs. To the intended end of using these data output, the system makes use of a linear syntactic structure and an vector of semantic attributes associated to the main verb at the sentence level.Evaluation of opinion and feeling is closely related to the semantic content of each sentence, assigned to two separate categories: objective vs. subjective. This differentiation is achieved through the search of some labels (eg. …