This code generates nearly believable titles for scientific papers based on a smoothed trigram model of all of the titles ever published on arxiv.org.

Usage

First download and unzip the text file containing the list of titles. Then run:

python ngram.py --build

After this first run, you only have to run

python ngram.py

and it will use the cached N-gram dictionary.

If you want to post to Twitter, register an app on the developer site (making sure that it has write access) and save a file called local.cfg with the format:

[twitter]
consumer_key: ...
consumer_secret: ...
user_key: ...
user_secret: ...

replacing the ... with the obvious things. Then call

python ngram.py --tweet

Licence

Licensed under the MIT license.

Copyright 2014 Dan Foreman-Mackey