Ask Me Anything: Dynamic Memory Networks for Natural Language Processing

http://photos2.meetupstatic.com/photos/event/6/1/2/b/600_445884875.jpeg

Speaker: Richard Socher is a Founder & CEO of a Deep Learning startup MetaMind (http://www.metamind.io). He has a PhD from Stanford CS department where his thesis advisors were Professors Chris Manning (http://nlp.stanford.edu/%7Emanning/) and Andrew Ng (http://www.cs.stanford.edu/people/ang/).

In spring 2015, he taught a class on Deep Learning for Natural Language Processing at Stanford.

Together with Chris Manning (http://nlp.stanford.edu/%7Emanning/) and Yoshua Bengio (http://www.iro.umontreal.ca/%7Ebengioy/yoshua_en/index.html), Richard is an author of Deep Learning Tutorial (http://www.socher.org/index.php/DeepLearningTutorial/DeepLearningTutorial) (Deep Learning for NLP, without Magic) at ACL 2012 and NAACL 2013.

List of Richard Socher's papers on Google scholar is here (http://scholar.google.com/citations?user=FaOcyfMAAAAJ&hl=en).

Abstract: Most tasks in natural language processing can be cast into question answering (QA) problems over language input. We introduce the dynamic memory network (DMN), a unified neural network framework which processes input sequences and questions, forms semantic and episodic memories, and generates relevant answers. Questions trigger an iterative attention process which allows the model to condition its attention on the result of previous iterations. These results are then reasoned over in a hierarchical recurrent sequence model to generate answers.

The DMN can be trained end-to-end and obtains state of the art results on several types of tasks and datasets: question answering (Facebook’s bAbI dataset), sequence modeling for part of speech tagging (WSJ-PTB), coreference resolution (Quizbowl dataset) and text classification for sentiment analysis (Stanford Sentiment Treebank). The model relies exclusively on trained word vector representations and requires no string matching or manually engineered features.

Link to Richard Socher profile: http://www.socher.org/

Location: Baidu Research

1195 Bordeaux Drive, Sunnyvale, CA

Please enter through front lobby. Meeting will be in Baidu Cafe.

Agenda:

6:00pm - Refreshments and networking

6:15-7:15pm - Dynamic Memory Networks for Natural Language Processing

7:15-8:00pm - Networking

This meeting is generously sponsored by Baidu Research.