Data Processing and its Impact on Linguistic Analysis

Abstract

The Saliba-Logea documentation project has been working toward a web-based text database with text-audio linkage and searchable annotations. In this article, I discuss the impact that the nature of data processing can have on linguistic analysis, and I demonstrate this on the basis of two research topics: the positioning of Postpositional Phrases and the distribution of plural markers. Saliba-Logea PPs can be ambiguous as to whether they belong to the preceding or following clause. To investigate whether there is a correlation between a PP’s position and its semantic role, text-only transcriptions turn out to be insufficient. The second question relates to the Saliba-Logea plural suffix, which originally occurred only on nouns with human referents. However, some speakers use it in novel contexts, and in order to investigate these extended uses and who drives them, access to metadata about the speakers is required. I show that text-audio linkage can be a prerequisite for analyzing syntactic constructions and that access to metadata can have a direct effect on the linguistic analysis.