Speech annotation of learner corpora

Grammatical and lexical development during study abroad: Research on a corpus of spoken L2 Chinese

“…The same 62 students' data were examined once right before or at the beginning of their program in China, and once after or at the end of their program in China, after one semester or 1 year of study in China. While cross‐sectional studies have been used widely in the field as a convenient proxy for developmental studies, development is best investigated truly longitudinally by following the same learners over a period of time to trace their performance in certain areas of the language, specifically using corpus linguistics research methods (e.g., Ballier & Martin, 2015; Meunier, 2015; Zhang & Tao, 2018). There is no widely agreed‐upon criteria as to how long is considered longitudinal in applied linguistics research, but one semester or one academic year of intensive study in a total immersion environment abroad, which is how data were collected in this study, should qualify (Ortega & Byrnes, 2008).…”

Directions for the future of technology in pronunciation research and teaching

“…Despite moves to address this lack of availability (Staples, 2015), little has changed yet (for examples, see French learner corpora at <http://www.flloc.soton.ac.uk/> and Spanish learner corpora at <http://www.splloc.soton.ac.uk/>. Ballier and Martin (2016) differentiate among three types of spoken learner corpora: mute corpora, which are transcriptions of spoken data; speaking corpora, which associate sound files with transcriptions of speech, such as the Speech Accent Archive, (Weinberger, 2017); and phonetic corpora, which align annotations with recorded speech signals. Often phonetic corpora include annotations at the levels of phrases, words, segments, and intonation, among others.…”

A corpus-assisted genre analysis of the Tunisian Lecture Corpus: An exploratory study

“…One major reason is that compiling spoken data is a daunting and expensive task, which is coupled with the sensitivity of the data under focus as well as with the current requirements for multimodal data. In fact, the use of muted spoken corpora (Ballier and Martin 2015), that is, transcripts of spoken language which are not distributed with audio and video files, does not account for the most basic and immediate context behind their production and their use fails to deliver a valid analysis and interpretation of the data (Deroey and Taverniers 2011). A second reason is the nature of the analyses that are conducted on academic lectures.…”