Morphosyntactic Analysis of Bangla Matrix Verbs: A Computational Perspective

This paper analyses verb morphology of Bangla matrix verbs or single predicates using a computational lens. It analyses the structure of matrix verbs in terms of tense-aspect-modality (TAM) markers as well as pronominal and valence marking, and proposes a conceptual Finite State Transducer (FST) that captures the rich inflectional feature-values in a concatenative manner. Often linguistic theoretic models and computational analyses are at loggerheads due to the units assumed, where linguistic theory is often grounded in psycholinguistic reality, while computational perspectives prioritise the optimisation of the implementation algorithm resulting in units that do not port over to linguistic theory. This paper analyses verb morphology in terms of TAM markers from a linguistic theoretic perspective and then extends the analysis by modeling a FST of said verbform with TAM and pronominal markers. The analysis is limited to single predicates or matrix verbs of Standard Colloquial Bangla (SCB), and can be extended to complex predicates by means of the concatenation of TAM markers on the vector or light verb, as the pole in a complex predicate remains static (Butt 2010, Masica 1993). The finite state transducer proposed here is a theoretical model, which can be implemented by researchers working with Bangla verb morphology and natural language processing (NLP) tools. The goal is to develop a more accurate and linguistically informed model of verb morphology for Bangla single and complex predicates that can serve both descriptive and applied purposes. This research work contributes to both the theoretical understanding of Bangla morphology and the practical advancement of NLP tools for low-resource languages.