Unsupervised Morphological Paradigm Completion

Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 6696–6707 July 5 - 10, 2020. c⃝2020 Association for Computational Linguistics 6696 Unsupervised Morphological Paradigm Completion Huiming Jin, Liwei Cai, Yihui Peng, Chen Xia Language Technologies Institute Carnegie Mellon University {huimingj,liweicai,yihuip,chenxia}@cs.cmu.edu Arya D. McCarthy Johns Hopkins University arya@jhu.edu Katharina Kann University of Colorado katharina.kann@colorado.edu Abstract We propose the task of unsupervised morpho- logical paradigm completion. Given only raw text and a lemma list, the task consists of gen- erating the morphological paradigms, i.e., all inflected forms, of the lemmas. From a nat- ural language processing (NLP) perspective, this is a challenging unsupervised task, and high-performing systems have the potential to improve tools for low-resource languages or to assist linguistic annotators. From a cognitive science perspective, this can shed light on how children acquire morphological knowledge. We further introduce a system for the task, which generates morphological paradigms via the following steps: (i) EDIT TREE retrieval, (ii) additional lemma retrieval, (iii) paradigm size discovery, and (iv) inflec- tion generation. We perform an evaluation on 14 typologically diverse languages. Our system outperforms trivial baselines with ease and, for some languages, even obtains a higher accuracy than minimally supervised systems.1 1 Introduction Morphologically rich languages express syntac- tic and semantic properties—like tense or case— of words through inflection, i.e., changes to the surface forms of the words. The set of all in- flected forms of a lemma—the canonical form—is called its paradigm. While English does not man- ifest a rich inflectional morphology, Polish verbs have around a hundred different forms (Sadowska, 2012), and Archi paradigms, an extreme example, can have over 1.5 million slots (Kibrik, 1977). Morphologically rich languages constitute a challenge for natural language processing (NLP) systems: because each lemma can take on a vari- ety of surface forms, the frequency of each indi- vidual inflected word decreases drastically. This yields problems for speech recognition (Creutz 1Our implementation is available under https://github. com/cai-lw/morpho-baseline. Sé�vigilante�y�confirma�las�otras�cosas�que están�para�morir�,�porque�no�he�hallado�tus obras�bien�acabadas�delante�de�Dios�. Acuérdate�,�pues�,�de�lo�que�has�recibido�y�oído�; guárdalo�y�arrepiéntete�,�pues�si�no�velas�vendré�sobre ti�como�ladrón�y�no�sabrás�a�qué�hora�vendré�sobre�ti�. El�vencedor�será�vestido�de�vestiduras�blancas�,�y�no borraré�su�nombre�del�libro�de�la�vida�,�y�confesaré�su nombre�delante�de�mi�Padre�y�delante�de�sus�ángeles�. El�que�tiene�oído�,�oiga�lo�que�el�Espíritu�dice�a�las iglesias�.�’�” »�Escribe�al�ángel�de�la�iglesia�en�Filadelfia�:�»�“�Esto dice�el�Santo�,�el�Verdadero�,�el�que�tiene�la�llave�de David�,�el�que�abre�y�ninguno�cierra�,�y�cierra�y�ninguno abre�: ... estar tener empezar pasar ... (1) EDIT TREES (2) NEW LEMMAS morir TAG1 mueres morir TAG2 mueren ...morir TAG3 morirás saber TAG1 sabes saber TAG2 saben ... (3) PARADIGM SIZE DISCOVERY (4) GENERATION morir saber vestir ... morir - mueres saber - sabes TAG1 Figure 1: Our unsupervised paradigm completion sys- tem, which takes raw text and a lemma list as inputs. We describe it in detail in §4. et al., 2007), parsing (Seeker and C¸ etino˘glu, 2015), and keyword spotting (Narasimhan et al., 2014), inter alia. For unsupervised machine translation, Guzm´an et al. (2019) encounter difficulties when translating into the morphologically rich languages Nepalese and Sinhalese. Children acquire morphological knowledge from raw utterances and, in particular, without access to explicit morphological information (Berko, 1958). Do they have an innate capacity that enables them to learn a language’s morphology? Or can morphol- ogy be learned in an unsupervised fashion? This question—in addition to practical considerations like benefits for the aforementioned …

Content truncated to 500 words