A Computational Model for the Linguistic Notion of Morphological Paradigm

Proceedings of the 27th International Conference on Computational Linguistics, pages 1615–1626 Santa Fe, New Mexico, USA, August 20-26, 2018. 1615 A Computational Model for the Linguistic Notion of Morphological Paradigm Miikka Silfverberg and Ling Liu and Mans Hulden Department of Linguistics University of Colorado first.last@colorado.edu Abstract In supervised learning of morphological patterns, the strategy of generalizing inflectional tables into more abstract paradigms through alignment of the longest common subsequence found in an inflection table has been proposed as an efficient method to deduce the inflectional behavior of unseen word forms. In this paper, we extend this notion of morphological ‘paradigm’ from earlier work and provide a formalization that more accurately matches linguist intuitions about what an inflectional paradigm is. Additionally, we propose and evaluate a mechanism for learning full human-readable paradigm specifications from incomplete data—a scenario when we only have access to a few inflected forms for each lexeme, and want to reconstruct the missing inflections as well as generalize and group the witnessed patterns into a model of more abstract paradigmatic behavior of lexemes. 1 Introduction Most work in phonology and morphology assumes that the ability to inflect previously unseen or unheard word forms is through a model of analogy (Blevins and Blevins, 2009) to known forms. In languages with poor inflectional morphology, this often manifests itself in the capacity to inflect previously unwitnessed lexemes, e.g. wug �→ wugs by analogy to some other lexeme, perhaps a noun like bug �→ bugs. A closely related problem, relevant for languages that have complex inflectional morphology and where a single part-of-speech can be inflected in perhaps thousands of ways, is the capacity to inflect a lexeme with a previously unheard combination of inflectional features. For example, Finnish nouns are inflected in 2,253 different ways,1 and no speaker will have witnessed all variants for any given noun. This capacity to fill in missing forms has been dubbed the paradigm cell filling problem (PCFP) (Ackerman et al., 2009). The frequency of references to the nebulous label ‘analogy’ and its responsibility for human perfor- mance in these tasks stands in contrast to the dearth of proposed exact formal or computational models of analogical word-formation in the literature. Even in second-language teaching material, speakers are as- sumed to possess the capacity to directly generalize a pattern from a collection of inflected forms. Table 1 shows a partial inflection table for the Spanish verb tostar (‘to toast’) given in the resource 501 Spanish Verbs (Kendris and Kendris, 2007). By contrast, the verb entry for colar (‘to strain’) contains no table at all, but simply provides a pointer to the inflection table for tostar. In other words, it is presupposed that, for example, given the analogy tostar:tuesto::colar:x, the astute reader can solve for x without undue problems, though no explicit procedure for doing so is given.2 One formalization of this implicit procedure is offered in Ahlberg et al. (2014) and Hulden (2014). In those works, a collection of inflected forms is first subjected to identifying the longest …

Content truncated to 500 words