Add wordfreq_data files.

Now the build process is repeatable from scratch, even if something goes
wrong with the download server.

.gitignore

@@ -6,5 +6,4 @@ dist
 pip-log.txt
 .coverage
 *~
-wordfreq_data/
 wordfreq-data.tar.gz

wordfreq_data/google/README.txt

@@ -0,0 +1,10 @@
+This data was compiled from the Google Books Ngram Viewer data, particularly
+the 2012 English dataset.
+
+The data is available from https://books.google.com/ngrams. The terms of use of
+this data are:
+
+"Ngram Viewer graphs and data may be freely used for any purpose, although
+acknowledgement of Google Books Ngram Viewer as the source, and inclusion of a
+link to http://books.google.com/ngrams, would be appreciated."
+

wordfreq_data/leeds/README.txt

@@ -0,0 +1,5 @@
+These wordlists come from the University of Leeds Centre for Translation
+Studies, and are provided for free under a Creative Commons Attribution
+license.
+
+For more information, see: http://corpus.leeds.ac.uk/list.html

wordfreq_data/luminoso/README.txt

@@ -0,0 +1,17 @@
+This directory contains two wordlists we've put together at Luminoso for our
+own purposes. You might find them useful as well.
+
+* `twitter-52M` collects the unigram word frequencies from 52 million tweets.
+  The words are not distinguished by language.
+
+* `multi` combines various sources of data in different languages, including:
+
+  * Google Books, for English
+  * A smaller corpus of tweets that supposedly come from English speakers
+    (there's still a lot of non-English text in there)
+  * the Leeds corpora for various languages (see `../leeds/README.txt`)
+
+We would like to release the tools that built `twitter-52M` as soon as they are
+less sloppy. `multi` is a dataset that is mainly relevant because it's the data
+we happen to already be using, but you might find it useful as well.
+