Learning New Facts From Knowledge Bases With Neural Tensor Networks and Semantic Word Vectors — mentioned by kmike84

Sun, 28 Dec 2014 14:13:53 GMT

Sure they can, see e.g. http://nlp.stanford.edu/pubs/SocherChenManningNg_NIPS2013.pd... or http://arxiv.org/pdf/1301.3618v2.pdf. Check the first paper - when working on this problem researchers from Stanford developed a way to measure the quality of an approach, evaluated the results on 30k+ relation examples from 2 different datasets and compared their algorithm with 4 other algorithms. The problem with cortical.io or numenta (both commercial companies) is that they don't compare their approaches with existing approaches and don't evaluate them on public datasets. And when people do such…

Paper and discussion history. A mention is not an endorsement.

Learning New Facts From Knowledge Bases With Neural Tensor Networks and Semantic Word Vectors · HackerPapers discussions

Learning New Facts From Knowledge Bases With Neural Tensor Networks and Semantic Word Vectors — mentioned by kmike84