The Paradigm Discovery Problem

Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 7778–7790 July 5 - 10, 2020. c⃝2020 Association for Computational Linguistics 7778 The Paradigm Discovery Problem Alexander Erdmann@,D Micha Elsner@ Shijie WuZ Ryan CotterellQ,6 Nizar HabashD @Ohio State University DNew York University Abu Dhabi ZJohns Hopkins University 6University of Cambridge QETH Zürich {ae1541,nizar.habash}@nyu.edu elsner.14@osu.edu shijie.wu@jhu.edu ryan.cotterell@inf.ethz.ch Abstract This work treats the paradigm discovery prob- lem (PDP)—the task of learning an inflec- tional morphological system from unannotated sentences. We formalize the PDP and develop evaluation metrics for judging systems. Us- ing currently available resources, we construct datasets for the task. We also devise a heuris- tic benchmark for the PDP and report empir- ical results on five diverse languages. Our benchmark system first makes use of word embeddings and string similarity to cluster forms by cell and by paradigm. Then, we bootstrap a neural transducer on top of the clustered data to predict words to realize the empty paradigm slots. An error analysis of our system suggests clustering by cell across different inflection classes is the most press- ing challenge for future work. Our code and data are available at https://github.com/ alexerdmann/ParadigmDiscovery. 1 Introduction In childhood, we induce our native language’s mor- phological system from unannotated input. For instance, we learn that ring and rang belong to the same inflectional paradigm. We also learn that rings and bangs belong to the same cell, i.e., they realize the same morphosyntactic properties 3.SG.PRES, but in different paradigms. Acquiring such paradigmatic knowledge enables us to pro- duce unseen inflectional variants of new vocabulary items, i.e. to complete morphological paradigms. Much work has addressed this task, which Ack- erman et al. (2009) call the paradigm cell filling problem (PCFP),1 but few have discussed inducing paradigmatic knowledge from scratch, which we call the paradigm discovery problem (PDP).2 1In the NLP literature, this task is called morphological reinflection or morphological inflection generation (Cotterell et al., 2016a); this is only a difference in nomenclature. 2Elsner et al. (2019) call the task the paradigm cell dis- covery problem; we drop cell to distinguish our task from As an unsupervised task, the PDP poses chal- lenges for modeling and evaluation and has yet to be attempted in its full form (Elsner et al., 2019). However, we contend there is much to be gained from formalizing and studying the PDP. There are insights for cognitive modeling to be won (Pinker, 2001; Goldwater, 2007) and intuitions on combat- ing sparse data for language generation (King and White, 2018) to be accrued. Unsupervised lan- guage processing also has natural applications in the documentation of endangered languages (Za- maraeva et al., 2019) where a lot of annotated data is never likely to exist. Our formalization of the PDP offers a starting point for future work on un- supervised morphological paradigm completion. Our paper presents a concrete formalization of the PDP. Then, as a baseline for future work, we in- troduce a heuristic benchmark system. Our bench- mark system takes an unannotated …

Content truncated to 500 words