Representing Complexity: A Numeric-Computational Approach to a Spoken Corpus on LGBTQ Discourse

Complexity theory is a common theoretical framework in physics to describe not only natural phenomena, but very recently also social and language behaviors (Baronchelli et al. 2012; Conte et al. 2012; Loreto et al. 2011; Gravino et al 2012; Goncalves & Perra 2015; Gong et al. 2012). The physics edge is not only based on a set of felicitous theoretical constructs with a close focus on real world phenomena, but also on a long established tradition in including computational investigation methods in quantitative study designs (Borgman 2015; Miller & Page 2007:5). How to gain an understanding of complexity is a non trivial question when tackling language issues. At an intuitive level, the very notion of style involves the use of fronted - in complexity theory parlance, emerging - linguistic devices being simultaneously chosen by speakers - the agents - on multiple rankscale levels (cf. Malmkjaer 2004; Holland 2014). All rankscale levels, however, aggregate at textual level to form higher level structures following arbitrary, non deterministic principles. In turn, the sum of all fronted devices from a locutionary point of view doesn't account for the overall force found on the illocutionary and perlocutionary side. Language and communication aren't simply 'complicated' systems: Removing or simplifying elements on any level of this hierarchy irreparably compromises not only the structure, but also the very meaning construction of a text. In addition, natural languages are pervaded by variation phenomena leading to instability and fragmentation in time, space, social layers and subject matter among others. Other clues of instability can intuitively be detected during the early stages of discourse formation: Speakers covering a common topic can learn from other speakers' utterances and adapt their linguistic choices over time, in a way that can be either optimal or not. This hints that language and discourse, understood as social and collaborative phenomena, can be modeled as a complex adaptive system (CAS). Far from acknowledging numeric and computational methods full accuracy in describing meaning construction and discourse behavior, this preliminary study focuses on the exploration of language phenomena (e.g. the Zipf's law) that may confirm the CAS status of the LGBTQ discourse, as represented by the transcriptions of live talks hosted by the TED Conferences. This analysis provides a survey of recurrent language patterns at word and phrase level, and their interaction both at text and corpus level. Choosing the LGBTQ discourse, an emerging one and one that is linked with taboo language and sensitive social issues, is key to assess in further studies if any adaptive behavior occurs in language phenomena over time and if there is any evidence of stylistic equilibria.