This is a guest post by Orin Hargraves, an independent lexicographer, language researcher, and past president of the Dictionary Society of North America. Orin is the author of several language reference books, including It’s Been Said Before: A Guide to the Use and Abuse of Clichés (Oxford) and Slang Rules!: A Practical Guide for English Learners (Merriam-Webster).
A few years ago I wrote about how collocations in fiction skew the statistics of collocations in a corpus because of their extremely frequent use; Ben Zimmer expanded on the idea in a later New York Times piece. In summary, the point is that a number of collocations would not be statistically significant were it not for their appearance in fiction. This is because writers of fiction—particularly writers of the amateur, unedited fiction that appears online—tend to reuse the same tropes and phrases so much that these effectively become clichés, formulaic ways of expressing the same (rather tired) ideas and events.
All of that came to light when I was working with the Oxford English Corpus, a well balanced and carefully curated corpus that, at the time, had about two billion words of English. These days I’m working with the enTenTen13 corpus, a web-crawled corpus of nearly 20 billion words, owned and made available by Sketch Engine. Sketch Engine’s web-crawler roves the Internet indiscriminately, pulling text from wherever it can be found. Like some grandmother aghast in Greenville, the web-crawler regularly comes upon sites with pornographic content. The difference between the grandmother and the web-crawler is that while she may avert her gaze in shock and dismay, the web-crawler grabs the text, parses and tags it, and adds it to the corpus. The result is that enTenTen13 houses a steaming, pulsating trove of pornographic writing.