I use corpus linguistics very much along the lines of John-Sinclair's
"Trust the text". I neither do corpus-assisted research (where a set theory and a set notion of language is simply proven by bringing in results from real (or natural) data nor is it corpus-based research (pretty similar to the above. However, findings are based on results found by corpora research). I undertake, I think, the more adventurous and less framework-bound version: corpus-led research. This means I approach data not with a fixed idea of what I want to find or how I would call collocational or colligational elements and structures. Too often, fixed notions and strict classifications have blinded researchers - and made them miss patterns. I believe that the natural occurring language itself provides the answers. Because of this, I am reluctant to use traditional terminology. It seems as if the most innocuous item (word) can have different functions and meanings and that is very much dictated by the item's surroundings. For this reason, "Lexical Priming" is a theory I find extremely helpful.
In my research (and, hopefully, very much the focus of my research
career) is the spoken word. This is in many ways the harder road to travel: transcribed spoken language is hard to come by and, where it exists, comes in the form of small corpora (very different to written corpora where researchers have now access to large chunks of the world wide web). Taping and transcribing one's own corpus is extremely time-consuming and tedious and, where third-party transcripts are being used, differing standards of transcription can render chunks of information useless. Speech is, however, our primary way of communication and, consequently, it corpora of spoken language should be able to reflect used speech at its purest. At the same time, spoken language is ideal for corpus-led research, as it is now widely accepted that people, when speaking, adhere to different grammatical rules than written text.
I compare material from my own corpus - over 100k words of spoken
Liverpool English, with two general corpora of spoken English from across the UK. The aim is to find whether a particular (geographical and / or socio-economic) speech group is primed to use certain key words and key phrases and ways that differ from an average found used across the country. I believe that this has been successful and that that I have proven that the English spoken in Liverpool is, sometimes subtly, sometimes more obviously, different in its use of the same lexical stock as found in the general corpora. I see this of evidence that members of one speech community (Liverpudlians) are primed to use certain items and phrases in ways that are either more or less strongly preferred by other English speakers in the UK.