Phonological Neighbourhood Density in Dutch Verbs: From Classification to Corpus Annotation

Phonological Neighbourhood Density (PND) quantifies how many phonologically similar verbs undergo the same type of stem alternation across tenses. In Dutch, for instance, blijken, lijken, and bijten form a neighbourhood through the transformation to bleek, leek, and beet in the simple past tense. PND has been widely studied in psycholinguistics, where denser neighbourhoods are linked to stronger analogical effects in lexical processing, morphological generalization, and language acquisition. A detailed PND classification of 1,452 Dutch verbs has been developed based on systematic vowel stem alternations between the simple present, simple past, and present perfect. Two types measures of PND are used: the PND-form measure counts how many verbs share the same phonological transformation from base to target tense (e.g., helpen–hielp, werpen–wierp), while the PND-class measure captures broader morphophonological generalizations, grouping verbs that follow the same stem alternation patterns across multiple tenses, e.g., simple present: lopen, simple past: liep, present perfect: gelopen. Building on this resource, we propose creating a phonological annotation layer for Dutch corpora within the CLARIN infrastructure (e.g., CGN, OpenSoNaR). Using standardized formats such as TEI or FoLiA, this layer will assign structured phonological categories to verb tokens and include relevant metadata. The resulting resource will support corpus-based research on language acquisition, phonology, morphology, and their interface with syntax and semantics, while also contributing to language technology applications and CLARIN’s mission to provide open, high-quality linguistic resources.