I found a great alternative: Microsoft Web N-Gram
It can be queried in different ways, including a straighforward GET call through the REST interface.
For instance, calling the URL:
http://weblm.research.microsoft.com/weblm/rest.svc/bing-body/apr10/1/jp?u={YOUR_TOKEN}&p=red+panda
returns
-9.005
which is the log likelihood of the phrase red panda.
Furthermore, it is handier than Google N-Grams, as for a given phrase it does not simply output its absolute frequency, but it can output its joint probability, conditional probability and even the most likely words that follow.
Disclaimer: I am not a Microsoft employee, I simply think that I just found an awesome service.
