IconNet and Pictorial Meaning Representation: Towards Language-Neutral Semantics
Xiao Zhang, Johan Bos
Center for Language and Cognition, University of Groningen
Current meaning representation formalism in NLP use natural language (usually English) to denote concepts, predicates, and relations. We investigate the possibilities of developing a WordNet-based ontology where synsets/concepts are represented by icons. The resulting IconNet resource is combined with graph-based semantic representations to capture the meaning of a sentence or short text, resulting in a multi-modal meaning representation combining textual and pictorial elements. We believe that the integration of these modalities significantly enhances the flexibility and applicability of meaning representations in diverse communicative contexts.