Although my friend and colleague, Jay McKinnon , has pointed out in no uncertain terms its limitations and weaknesses, I couldn't resist trying out the Google N-Gram Viewer. For those of you even farther behind on the digital humanities bandwagon than I am, this application allows one to search for keywords in the Google Books database and generate a graph of their frequencies (displayed as proportion of the corpus in that particular year). See, for example, the relative frequencies of "nerd," "geek," "dork," and "dweeb" between 1975 and 2007:
This graph, despite methodological differences, seems to mirror the trends that I've found tabulating the frequency of "nerd" and "geek" in newspaper articles in the LexisNexis Academic database.
In my own work, I'm still sorting out false positives for "geek" as in "Carnival chicken-biter," but both sources indicate a significant increase over the last 35 years or so. In both cases, "nerd" is originally the more common term, but "geek" surpasses it in the early-2000s. My best guess is that, while both terms are used relatively interchangeably with respect to media-related practices and fandoms, "geek" has been more central among technical fields, such that the dot-com boom/bust puts it over the top.