Camilleri’s Sicilian: From Stopwords To Topic Modelling

Andrea Camilleri represents a peculiar case in Italian narrative for the conscious use of a mixed language, which combines Italian and Sicilian. This linguistic choice generates a unique lexicon, distant from the conventions of standard Italian and only partially attributable to codified Sicilian dialect; this represents a challenge for computational linguistics. In this work we present the construction of a stop-words list dedicated to the Sicilian lexicon present in Camilleri’s works, aimed at the automatic analysis of topics. Starting from a corpus that collects the 70 works of Camilleri published by Sellerio, Sicilian lemmas were identified through a manual annotation procedure that combined native-speaker expertise with systematic consultation of the Sicilian vocabulary of the Centro di Studi Filologici e Linguistici Sicliani and the Dizionario Camilleriano/Italiano by Mario Genco. The objective of the constructed list is to distinguishing dialectal forms from Italian lexicon. Subsequently, stopwords were identified and words with semantic relevance were selected. The result is a Camilleri’s Sicilian lexicon, which collects morphosyntactic information. This resource constitutes a useful contribution both to the automatic processing of natural language in the context of the Sicilian dialect and to authorial characterisation and the study of linguistic variability in literary dialectal contexts. As a first application, we estimated a Structural Topic Model incorporating this dialect-specific information obtaining interpretable topics that align with the series. Our results show that a targeted author-specific dialectal annotation with structural topic models and document-level covariates offers a useful starting point for studying other Italian regional and dialectal literary varieties where standard-language tools are inadequate.