A Corpus Too Small: Uses of Text Data in a Hupa-English Bilingual Dictionary

Although corpus-driven methods have led to a revolution in the way lexicographers of some languages approach their work, text corpora for many less-studied languages are too small for such methods to be used reliably. Hupa, a Native American language of northwestern California, is one such language. Nonetheless, the Hupa Online Dictionary and Texts website relies heavily on its small text corpus to support development of the dictionary component. The corpus is especially important as a way to address Hupa’s complex and productive polysynthetic morphology, both derivational and inflectional, with words attested in the corpus providing the empirical basis for creating new entries and expanding the coverage of existing ones. It also provides a ready source of example sentences in context, figurative uses of language that might not come to light through elicitation, and aspects of linguistic variation that dictionary normalization tends to obscure. Thus, while corpus-driven lexicography may not be a realistic possibility at this point, corpus-based lexicography (Tognini-Bonelli 2001) is certainly within reach.