Source-Path-Goal: Investigating the Cross-Linguistic Potential of Frame-Semantic Text Analysis
Oliver Czulo, Gerard de Melo · it - Information Technology · 2012
Abstract The ability to analyse natural language semantically has been a long-standing goal in information technology. In the last few decades, the increased availability of annotated text collections and new statistical methods have led to great advances in this area. Frame-semantic annotation, as exemplified by the FrameNet project, describes the meaning of a sentence in terms of the involved cognitive scenarios and the roles of individual participants (e. g. the buyer and the seller in a commerce scenario, but also abstract entities like source, path, and goal in a travel scenario). This contribution studies the cross-linguistic potential of FrameNet from multiple perspectives. It investigates a statistical method to determine frame-semantic relatedness of individual words across languages. It also shows how commonalities and differences between languages in expressing states of affairs can be analysed and extracted from annotated parallel corpora.