Annotation of metaphorical expressions in the Basic Corpus of Polish Metaphors

Abstract

This paper presents a corpus of Polish texts annotated with metaphorical expressions. It is composed of two parts of comparable size, selected from two subcorpora of the Polish National Corpus: the subcorpus manually annotated on morphosyntactic level, named entities level etc., and the Polish Coreference Corpus, with manually annotated mentions and the coreference relations between them, but automatically annotated on the morphosyntactic level (only the second part is actually annotated). In the paper we briefly outline the method for identifying metaphorical expressions in a text, based on the MIPVU procedure. The main difference is the stress put on novel metaphors and considering neologistic derivatives that have metaphorical properties. The annotation procedure is based on two notions: vehicle – a part of an expression used metaphorically, representing a source domain and its topic – a part referring to reality, representing a target domain. Next, we propose several features (text form, conceptual structure, conventionality and contextuality) to classify metaphorical expressions identified in texts. Additionally, some metaphorical expressions are identified as concerning personal identity matters and classified w.r.t. their properties. Finally, we analyse and evaluate the results of the annotation.

Resources

Details