DataTaunew | comments | leaders | submitlogin
Statistical Dependency Parsing using SVM in Python by @rj_here (github.com)
2 points by rjbullet 2971 days ago | 3 comments


1 point by gaina 2971 days ago | link

Training with so many features using support vector machines seems hard. How much time did it take to train?

-----

1 point by rjbullet 2971 days ago | link

You are right, it takes a really long time. To handle this the original author of the paper suggested using different SVM's based on the POS tag of the preceding word to make it practical. Even after that it took about 4-5 hours to train models for POS tags like NN on an aws machine with 32 gb ram. If you want to try the trained models are available at :http://bit.ly/nlp-cs6741-full.

-----

1 point by rjbullet 2971 days ago | link

The report with results of experimentation on biomedical text and spanish data is available here: http://bit.ly/nlp-cs6741

-----




RSS | Announcements