An Enhanced Pipeline for the Manzini-Savoia Dialect Corpus

Abstract

This paper presents a semi-automatic workflow for enriching the Manzini–Savoia Corpus (MSC) of Italian dialects with extended glosses, normalized transcriptions, and projected morpho-syntactic annotations. While the MSC is a unique resource for Romance microvariation, its partial glossing and phonetic transcription in the International Phonetic Alphabet (IPA) pose major challenges for computational processing. We introduce a pipeline for gloss coverage expansion and reliable morpho-syntactic annotation combining rule-based and data-driven components, which includes: (i) automatic completion of truncated verbal paradigms; (ii) hybrid lexical alignment between dialectal tokens and Italian glosses, integrating per-region lexical priors with a dynamic programming alignment algorithm; and (iii) projection-based morpho-syntactic tagging from aligned glosses. The proposed methods offer a reproducible framework for extending partially glossed dialect corpora and contribute new annotated data for research in computational dialectology and cross-variety language modeling.

Resources

Details