Can recursive neural tensor networks learn logical reasoning? [pdf]

http://arxiv.org/pdf/1312.6192.pdf

by mutor • 11 years ago

38 3 11 years ago

soup10• 11 years ago

So the paper says yes, but offers no analysis on why it's potentially better or different from other models of logical reasoning?

blake_himself• 11 years ago

For NLP. Presumably you could extract it from sentences with techniques like Socher's:

http://www.socher.org/index.php/Main/ParsingNaturalScenesAnd...

There was an example in Socher's paper of a tree of logic - ands, ors, nots (not shown on that site, but in the paper) - which logic, he showed his RNN technique can represent. The OP paper I imagine is more along those lines. Being able to capture full logic like that gets you that much closer to being able to extract the full meaning from language, not just words that provide a general flavor.

JadeNB• 11 years ago

My usual plea for abstracts over PDFs: http://arxiv.org/abs/1312.6192 .

Can recursive neural tensor networks learn logical reasoning? [pdf]

Related Stories