Linguistic Linked Data for Sentiment Analysis

In this paper we describe the specification of a model for the semantically interoperable representation of language resources for sentiment analysis. The model integrates ‘lemon’, an RDF-based model for the specification of ontology-lexica (Buitelaar et al. 2009), which is used increasingly for the representation of language resources as Linked Data, with 'Marl', an RDF-based model for the representation of sentiment annotations (Westerski et al., 2011; Sánchez-Rada et al., 2013). In the EuroSentiment project, the lemon/Marl model will be used to represent lexical resources for sentiment and emotion analysis such as SentiWordNet (Baccianella et al. 2010) and WordNet Affect (Strapparava and Valitutti 2004), as well as other language resources such as sentiment annotated corpora, in a semantically interoperable way, using Linked data principles. The representation of WordNet resources in lemon depends on a straightforward conversion of the WordNet data model, but importantly we introduce the use of URIs to uniquely and formally define structure and content of this WordNet based language resource. URIs are adopted from existing Linked Data resources, thereby further enhancing semantic interoperability. We further integrate a notion of domains into this representation in order to enable domain-specific definition of polarity for each lexical item. The lemon model allows for the representation of all aspects of lexical information, including lexical sense (word meaning) and polarity, but also morphosyntactic features such as part-of-speech, inflection, etc. This kind of information is not provided by WordNet Affect but will be available from other language resources, including those available at EuroSentiment partners that can be