Remote Elicitation of Inflectional Paradigms to Seed Morphological Analysis in Low-Resource Languages

Remote Elicitation of Inflectional Paradigms to Seed Morphological Analysis in Low-Resource Languages John Sylak-Glassman*, Christo Kirov*, David Yarowsky** *Center for Language and Speech Processing **Department of Computer Science Johns Hopkins University Baltimore, MD 21218, USA jcsg@jhu.edu, ckirov@gmail.com, yarowsky@jhu.edu Abstract Structured, complete inflectional paradigm data exists for very few of the world’s languages, but is crucial to training morphological analysis tools. We present methods inspired by linguistic fieldwork for gathering inflectional paradigm data in a machine-readable, interoperable format from remotely-located speakers of any language. Informants are tasked with completing language-specific paradigm elicitation templates. Templates are constructed by linguists using grammatical reference materials to ensure completeness. Each cell in a template is associated with contextual prompts designed to help informants with varying levels of linguistic expertise (from professional translators to untrained native speakers) provide the desired inflected form. To facilitate downstream use in interoperable NLP/HLT applications, each cell is also associated with a language-independent machine-readable set of morphological tags from the UniMorph Schema. This data is useful for seeding morphological analysis and generation software, particularly when the data is representative of the range of surface morphological variation in the language. At present, we have obtained 792 lemmas and 25,056 inflected forms from 15 languages. Keywords: Low-Resource Languages, Morphology, UniMorph Schema, Seed Corpus, Crowdsourcing, Linguistic Fieldwork 1. Introduction For most of the world’s languages, no structured, com- plete inflectional paradigms in a machine-readable format are available for human language technology (HLT) ap- plications. These paradigms are necessary as seed data for downstream tasks, especially morphological analysis and generation. Increased global internet access has made it possible to gather inflectional paradigm data directly from remotely-located speakers of many of the world’s lan- guages. We present methods for data collection grounded in the practice of linguistic fieldwork. The methods devel- oped thus far have been used to gather a corpus of inflec- tional paradigm data for 15 languages, with materials ready for eliciting data from speakers of 33 additional languages.1 The methods we have developed entail first constructing detailed elicitation templates in which speakers supply de- sired inflectional forms (§2.). These templates include: 1) A prompt in a high-resource world language (e.g. English) that speakers can translate into the target low-resource lan- guage (e.g. Uzbek), 2) a short description of the desired inflectional form in pedagogical or linguistic terminology (e.g. ‘aorist’), and 3) a detailed specification of the inflec- tional form in terms of the UniMorph Schema, introduced by Sylak-Glassman et al. (2015a, 2015b).2 These elicita- 1These 15 languages are listed in Table 1, and the additional 33 languages include: Acholi, Armenian, Balochi, Burmese, Cambo- dian, Coptic, Dinka, Gujarati, Hausa, Hindi, Kikuyu, Kongo (Fi- oti), Kurdish, Lao, Lingala, Luganda, Maguindanao, Malay, Mon- golian, Nahuatl, Nubian, Nuer, Pashto, Samoan, Somali, Span- ish, Surubu (Fiti), Swahili, Tausug, Thai, Vietnamese, Wolof, and Zande. 2The name used in those works for this schema, the Universal Morphological Feature Schema, is deprecated in favor of the name UniMorph Schema. tion templates are distributed to remotely-located speakers, who may be either professional …

Content truncated to 500 words