-
Notifications
You must be signed in to change notification settings - Fork 357
Changelog
Abandoned Neo4j after yet another out-of-memory crash. Using MongoDB instead.
Concepts may now have extra slashes in them, forming ambiguous and unambiguous concepts.
Ambiguous concepts have the form /concept/lang/NAME.
Disambiguated concepts have the form /concept/lang/NAME/POS/SENSE. Here, POS is a one-letter part of speech representing the role of the head word of the concept (n=noun, v=verb, a=adjective). Only nouns, verbs, and occasionally adjectives should be head words. SENSE is a hopefully-unique string identifying the sense of the word, which is probably going to differ based on where we get the word senses from.
For example, Wiktionary may give us /concept/en/sink/n/basin versus /concept/en/sink/v/submerge. /concept/en/sink will have edges of type "sense" pointing to both.
Concept normalization now depends on WordNet, which we access through NLTK.
If you don't yet have NLTK, download it from http://nltk.googlecode.com/files/nltk-2.0.1rc1.zip. Unzip it and run python setup.py install. (Put sudo before that if you're installing into a global Python environment on Unix.)
- Changed URIs so they can contain arbitrarily nested lists.
- Added sources and conjunctions as new node types.
- Added the "normalized" relation, which points from an assertion (or concept) to a simplified/normalized version of it.
- Began importing data from the pet game.
Starting points
Reproducibility
Details