This is an outstanding 2008 NLP paper and work, they elaborate a distributional language model with word embeddings learnt with a deep net in a system called SENNA that has a surprisingly small c source code available in adition to a pretrained english model allowing one to do semantic role labeling that gives you a shallow semantic parsing just almost after unpacking once one learns how they display the results, and the computation goes at neural net speed, with competitive precision/recall with other reserach methods.
2 comments
[ 3.1 ms ] story [ 13.4 ms ] thread