Construction of the Gmane corpus for examining the diffusion of lexical innovations

Large-scale linguistic corpora, complete with information about speakers’ social networks as well as demographic and temporal information, allow for empirical validation of complex theories about the social interactions and linguistic properties leading to large-scale language change. We present ongoing work on the diffusion of lexical innovations using a corpus we have compiled from the Gmane electronic mailing list archive, a publicly available dataset of 13,494 mailing lists and 117,606,370 messages to date. Focusing initially on a single list, we derive a social network for actor-speakers, give lexical and network statistics, and empirically categorize tie strength across speakers. Initial explorations of the Gmane corpus suggest suitability for research on language change.