The streaming access to the Google ngram data.
The data is so big, that storing it is almost impossible. However, sometimes you need an aggregate data over the dataset. For example to build a co-occurrence matrix.
This package provides an iterator over the dataset stored at Google. It decompresses the data on the fly and provides you the access to the underlying data.
- Download ngrams of various length and languages.
- Access to part of ngrams, e.g. ones that start with an ‘a’.
pip install google-ngram-downloader
The command line tool
It also provides a simple command line tool to download the ngrams called google-ngram-downloader. Refer to the help to see available actions:
google-ngram-downloader help usage: google-ngram-downloader <command> [options] commands: cooccurrence Write the cooccurrence frequencies of a word and its contexts. download Download The Google Books Ngram Viewer dataset version 20120701. help Show help for a given help topic or a help overview. readline Print the raw content.
Example use of the API
>>> from google_ngram_downloader import readline_google_store >>> >>> fname, url, records = next(readline_google_store(ngram_len=5)) >>> fname 'googlebooks-eng-all-5gram-20120701-0.gz' >>> url 'http://storage.googleapis.com/books/ngrams/books/googlebooks-eng-all-5gram-20120701-0.gz' >>> next(records) Record(ngram=u'0 " A most useful', year=1860, match_count=1, volume_count=1)
- Citation information.
- Tests are correctly packaged for a release.
- Added ‘indices’ keyword. Thanks to neocortex.
- Added ‘language’ flat. Thanks to Ray Powell (rpowellgit).
- Non-unique contexts are taken into account inside of an ngram.
- The cooccurrence command does not perform any ngram modification.
- download, readile and cooccurrence subcommands.
- readline_google_store transforms lines to Record in several processes.