Considering the storage, management, and processing of spontaneous speech corpora: New methods and new findings.

Abstract

Recordings of speech, both spontaneous and nonspontaneous, comprise the backbone of acoustic phonetic, speech science, and much general linguistic research. The creation of these speech recordings (e.g., recording equipment and its use) and their analysis (e.g., appropriate statistical methods) tend to be thoroughly treated in methodology sections of papers and in specialized methodology textbooks. However, the storage, management, and preservation of these resources are rarely discussed in the academic literature, though these practices influence both the short‐term and long‐term usability of these resources. This talk addresses the related questions of how we might best manage large and growing collections of audio data and how we can leverage new technologies so that our data archives are not just usable, but maximally useful. Examples are provided from two web‐based archiving projects, the Sociolinguistic Archive and Analysis Project (SLAAP) and the Online Speech/Corpora Archive and Analysis Resource, which feature organization and analysis tools that enhance the overall usefulness of the archived recordings. To exemplify some substantive outcomes of these projects, findings from a corpus sociophonetic study of speech rate and pause variation in American English, made possible by SLAAP’s software, are also discussed.