Towards a Lexicon-grammar based Framework for NLP: an Opinion Mining Application

Proceedings of Recent Advances in Natural Language Processing, pages 160–167, Hissar, Bulgaria, Sep 7–9 2015. Towards a Lexicon-grammar based Framework for NLP an Opinion Mining Application Annibale Elia, Serena Pelosi, Alessandro Maisto and Raffaele Guarasci Department of Political, Social and Communication Science University of Salerno {elia,spelosi,amaisto,rguarasci}@unisa.it Abstract The present research exploits the large amount of linguistic resources developed into the Lexicon-grammar paradigm in the domain of the Opinion Mining. Grounded on the Semantic Predicates theory, the proposed system is able to automatically match the syntactic structures selected by special classes of verbs, indicating positive or negative Sentiment, Opinion or Physi- cal acts, with the semantic frames evoked by the same lexical items. This methods has been tested on a large dataset com- posed of short texts, such as tweets and news headings. 1 Introduction In our research we propose a computational use of the Lexicon-grammar (LG) theories in the domain of the Opinion Mining. We take advantage of both the huge amount of lin- guistic facts, accurately formalized and described in the LG paradigm, and from the possibility to apply and test them on big data. The purpose is to build a fine grained Information Extraction tool able to locate meaningful information in raw texts and to characterize them with thorough semantic descriptions. According with the Semantic Predicates theory (Gross, 1981), it has been possible to perform a matching between the definitional syntactic struc- tures, attributed to each class of verbs, and the se- mantic information we attached in the database to every lexical entry. This way we could create a strict connection be- tween the arguments, selected by a given Predi- cate listed in our tables, and the actants involved into the same verb’s Semantic Frame (Fillmore, 1976; Fillmore and Baker, 2001; Fillmore, 2006). Thanks to our LG-based linguistic rules, anchored on the Semantic Predicates, we started with this research the development of an NLP framework that, on the base of sophisticated syntactic and se- mantic analyses, extracts real text occurrences and labels them with the semantic roles involved in ev- ery matched sentence. This ambitious work, that in this preliminary stage focused just on the predicates indicating senti- ments, opinions and physical acts, intends, in fu- ture works, to become a larger development of an LG-based Italian cross-platform open library for various kind of linguistic analyses. We excluded from this work the Transfer, the Spa- tial and the also the Psychological Predicates, be- cause they have been already tested on different kinds of raw data with satisfactory results (Vietri, 2014; Elia et al., 2010; Elia and Vietri, 2010; Elia et al., 2013; Maisto and Pelosi, 2014b). 2 Theoretical Background The Lexicon-grammar (LG), the method and the practice of formal description of the natural lan- guage, introduced important changes in the way in which the relationship between lexicon and syntax was conceived. (Gross, 1971; Gross, 1975). In the LG theoretical framework the minimum discourse units endowed with meaning are the whole nuclear sentences, generally anchored on the verbs, which …

Content truncated to 500 words