Design, Transcription, and Analysis of the Tunisian Lecture Corpus (TLC): Context related Issues

The importance of context with regard to corpus studies is reflected in the move from general to specialized corpora, particularly those accounting for different language varieties and/or disciplines. In terms of methodology, however, progress has been slower as discourse and genre approaches to corpus analysis remain insufficient and particularly so when it comes to the analysis of oral corpora. Moreover, the collection and transcription of a spoken corpus are complex and time-consuming processes, as the right balance needs to be found between authenticity and practicality. On the one hand, both processes seek to maximize the contribution of context in the interpretation and exploitation of corpus data for analytical and pedagogical purposes. On the other hand, designing an oral corpus is a time consuming and expensive endeavor. Consequently, contextual aspects that are important for data interpretation may not be collected. Transcripts should also be machine-readable for different analytical procedures to be performed, which may compromise their authenticity. In highlighting these issues, the present paper discusses the design, transcription, and analysis of TLC, a nonnative English corpus of lectures in three disciplines, and concludes that attempts to preserve contextual features may not be fully successful in capturing the dynamic nature of discourse.