Paradigm Completion for Derivational Morphology

Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing, pages 714–720 Copenhagen, Denmark, September 7–11, 2017. c⃝2017 Association for Computational Linguistics Paradigm Completion for Derivational Morphology Ryan Cotterell@ Ekaterina VylomovaH Huda Khayrallah@ Christo Kirov@ David Yarowsky@ @Center for Language and Speech Processing, Johns Hopkins University, USA HDepartment of Computing and Information Systems, University of Melbourne, Australia ryan.cotterell@jhu.edu evylomova@student.unimelb.edu.au {huda,ckirov,yarowsky}@cs.jhu.edu Abstract The generation of complex derived word forms has been an overlooked problem in NLP; we fill this gap by applying neu- ral sequence-to-sequence models to the task. We overview the theoretical mo- tivation for a paradigmatic treatment of derivational morphology, and introduce the task of derivational paradigm completion as a parallel to inflectional paradigm com- pletion. State-of-the-art neural models, adapted from the inflection task, are able to learn a range of derivation patterns, and outperform a non-neural baseline by 16.4%. However, due to semantic, historical, and lexical considerations involved in deriva- tional morphology, future work will be needed to achieve performance parity with inflection-generating systems. 1 Introduction Unlike inflectional morphology, which produces grammatical variants of the same core lexical item (e.g., take�→takes), derivational morphology is one of the key processes by which new lemmata are created. For example, the English verb corrode can evolve into the noun corrosion, the adjective corro- dent, and numerous other complex derived forms such as anticorrosive. Derivational morphology is often highly productive, leading to the ready cre- ation of neologisms such as Rao-Blackwellize and Rao-Blackwellization, both originating from the Rao-Blackwell theorem. Despite the prevalence of productive derivational morphology, however, there has been little work on its generation. Com- monly used derivational resources such as Nom- Bank (Meyers et al., 2004) are still finite. Moreover, the complex phonological and historical changes (e.g., the adjectivization corrode�→corrosive) and affix selection (e.g., choosing between English de- verbal suffixes -ment and -tion) make generation of derived forms an interesting and challenging problem for NLP. In this work, we show that viewing derivational morphological processes as paradigmatic may be fruitful for generation. This means that there are a number of well-defined form-function pairs as- sociated with a core word. For example, a typical English verb may have five forms in its inflectional paradigm, corresponding to its base (take), past tense (took), past participle (taken), progressive (taking) and third-person singular (takes) forms. These forms are related by a consistent set of re- lations, such as affixation. Similarly, a verb may have several slots in its derivational paradigm: The form take has the agentive nominalization taker, and the abilitative adjectivization takable. Note there are also consistent patterns associated with each derivational slot, e.g., the -er suffix regularly produces the agentive. Exploiting this paradigmatic characterization of derivational morphology allows us to create a sta- tistical model capable of generating derivationally complex forms. We apply state-of-the-art models for inflection generation, which learn mappings from fixed paradigm slots to derived forms. Em- pirically, we compare results for two models on the new task of derivational paradigm completion: a neural sequence-to-sequence model and a …

Content truncated to 500 words