Automatic extraction of opera character characteristics through lexical-syntactic patterns

Abstract

The paper presents the experimentation of rules based on syntactic patterns for the extraction of interpersonal relationships between characters from the opera repertoire of the 18th century. The study illustrates the application of this unsupervised approach to identify this kind of relationships described within the captions accompanying the lists of characters in the librettos. The results demonstrate the effectiveness of the proposed solution in extracting relations defined through formal ontologies with a precision that allows them to be included in a domain knowledge base without further supervision. The experimentation contributes to the elaboration of a formal model for the description of the features of opera characters required for the reconstruction of their profiles. Given the size of the repertoire under examination and the number of characters, the use of automated analysis offers the researcher a useful tool to support the critical analysis of the sources. The ability to reconstruct the network of relationships and the features of the characters is in turn preparatory to the use of the character as a dimension of analysis to reconstruct the complex tradition of this kind of texts.

Keywords

• information extraction
• nlp
• characters
• opera
• patterns