Dialectal varieties pose major challenges for NLP resource development, especially when annotation frameworks are organized around standardized language specifications. In Universal Dependencies (UD), dialects without independent ISO codes are subsumed under the corresponding standard language and inherit its language-level documentation, validator settings, and grammatical inventories. This paper examines Griko, a Greek variety spoken in southern Italy that developed in relative isolation from the Modern Greek dialect continuum while remaining in long-term contact with local Italo-Romance varieties. We assess the consequences of this organizational structure through controlled parsing experiments comparing intra-dialectal training, cross-dialectal transfer from Standard Modern Greek (SMG), script-controlled transfer using romanized SMG, and contact-related cross-lingual transfer from Italian. Our results show that, before romanization, the Italian model even surpasses SMG on several UD metrics and that, although romanization substantially improves SMG-based transfer, performance still remains far below the intra-dialectal baseline. We argue that this persistent gap reflects the interaction between structural divergence and language-level validation constraints, a phenomenon we term ISO-based validation coupling. Through analyses of auxiliary systems, voice marking, and progressive constructions, we show how standard-centric validation architectures can constrain the representation of dialect-specific grammar. More broadly, the Griko case highlights the limitations of language-centric organization in UD and underscores the need for variety-sensitive mechanisms when extending universal annotation frameworks to structurally divergent dialects.

Most work on part-of-speech (POS) tagging is focused on high resource languages, or examines low-resource and active learning settings through simulated studies. We evaluate POS tagging techniques on an actual endangered language, Griko. We present a resource that contains 114 narratives in Griko, along with sentence-level translations in Italian, and provides gold annotations for the test set. Based on a previously collected small corpus, we investigate several traditional methods, as well as methods that take advantage of monolingual data or project cross-lingual POS tags. We show that the combination of a semi-supervised method with cross-lingual transfer is more appropriate for this extremely challenging setting, with the best tagger achieving an accuracy of 72.9%. With an applied active learning scheme, which we use to collect sentence-level annotations over the test set, we achieve improvements of more than 21 percentage points.