The challenges and benefits of annotating oral bilingual corpora

Abstract

This article describes efforts to collect, process, and automatically annotate a corpus of Spanish as spoken in Texas. It elaborates the protocols for the development of the corpus and the procedures for automatic annotation, illustrating the common pitfalls to language identification in bilingual corpora and potential methods for circumventing them. The benefits of a comparative corpus approach to contact varieties is illustrated by a case study of a putative verbal calque from the Spanish in Texas data. It is demonstrated that the relative frequency of the verb is much higher than in its source Mexican variety and that the verb selects different complements in Texas than it does in other varieties. The article concludes with a discussion of how computational tools might be fruitfully exploited to resolve long-standing debates about language variation in contact settings.