-
Notifications
You must be signed in to change notification settings - Fork 201
Adds log-linear weighting of features for disambiguation #390
base: development
Are you sure you want to change the base?
Adds log-linear weighting of features for disambiguation #390
Conversation
echo "Downloading ranklib..." | ||
mkdir -p $BASE_WDIR/ranklib/ | ||
cd $BASE_WDIR/ranklib/ | ||
curl -L -o RankLib-2.1-patched.jar http://downloads.sourceforge.net/project/lemur/lemur/RankLib-2.1/RankLib-2.1-patched.jar?r=http%3A%2F%2Fsourceforge.net%2Fprojects%2Flemur%2Ffiles%2Flemur%2FRankLib-2.1%2F&ts=1439317425&use_mirror=skylink |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I was wondering if we could add RankLib as a maven dependency ? There is a suggestion on how to do it here http://sourceforge.net/p/lemur/discussion/ranklib/thread/a45e2a7c/?limit=25.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Yeah, might be nice to do that! On the other hand it's only a training-time dependency, right?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
True, but I find it a bit ugly downloading the jar on training time.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
We do the same type of thing for pignlproc and other training dependencies, I think I kind of imitated that
cab0bc2
to
132088b
Compare
…eeze 0.10 dependency
…eeze 0.10 dependency
…working and we can run disambiguations. Also using camel case now ;)
…ring VectorContextSimilarity accordingly
…nted vector store building in createspotlightmodel
…ript isn't fully running yet
…t to be worked out here. Also added a standalone w2v model file converter.
12eea64
to
44fdfb7
Compare
…re-introduces lost changes from the last commit of my original PR branch.
Re-introduce lost changes from the last commit of my original PR branch.
This is a version of #373 that tries to remove some duplicated commits. The original PR adds Log Linear Model based weighting to the disambiguation procedure. It also adds some new features, such as a bias term and a lexical similarity measure of the surface form and candidate entity name.