Deep Learning for NLP resources

Introductory and state of the art resources for NLP sequence modeling tasks like dialog.

##Machine Learning: Neural Networks, RNN, LSTM Coursera: Machine Learning
Andrew Ng
Introductory course for linear regression, logistic regression, and neural networks.
Also covers support vector machines, k-means, etc.

Cousera: Neural Networks
Geoffrey Hinton
Covers a variety of topics: Neural nets, RNNs, LSTMs.

Neural Networks (Playlist)
Covers : Neural nets, CRFs, RBMs, Autoencoder, DBNs, etc.

Machine Learning for Developers
Intro to basic ML concepts for developers.

Deep Learning (Book)
Yoshua Bengio
Advanced book about deep learning.

A few useful things to know about machine learning
Pedro Domingos

Understanding LSTM Networks
Blog post by Chris Olah.

Deep Learning for NLP

Stanford Natural Language Processing
Intro NLP course with videos. This has no deep learning. But it is a good primer for traditional nlp.

Stanford CS 224D: Deep Learning for NLP class
Richard Socher. (2015) Class with videos, and slides.

A Primer on Neural Network Models for Natural Language Processing
Yoav Goldberg. October 2015. No new info, 75 page summary of state of the art.

Word Vectors

Resources about word vectors, aka word embeddings, and distributed representations for words.
Word vectors are numeric representations of words that are often used as input to deep learning systems. This process is sometimes called pretraining.

Efficient Estimation of Word Representations in Vector Space
[Distributed Representations of Words and Phrases and their Compositionality] (http://papers.nips.cc/paper/5021-distributed-representations-of-words-and-phrases-and-their-compositionality.pdf)
Mikolov et al. 2013.
Generate word and phrase vectors. Performs well on word similarity and analogy task and includes Word2Vec source code Subsamples frequent words. (i.e. frequent words like "the" are skipped periodically to speed things up and improve vector for less frequently used words)
Word2Vec tutorial in TensorFlow

Deep Learning, NLP, and Representations
Chris Olah (2014) Blog post explaining word2vec.

GloVe: Global vectors for word representation
Pennington, Socher, Manning. 2014. Creates word vectors and relates word2vec to matrix factorizations. Evalutaion section led to controversy by Yoav Goldberg
Glove source code and training data

Thought Vectors

Thought vectors are numeric representations for sentences, paragraphs, and documents. The following papers are listed in order of date published, each one replaces the last as the state of the art in sentiment analysis.

Recursive Deep Models for Semantic Compositionality Over a Sentiment Treebank
Socher et al. 2013. Introduces Recursive Neural Tensor Network. Uses a parse tree.

Distributed Representations of Sentences and Documents
Le, Mikolov. 2014. Introduces Paragraph Vector. Concatenates and averages pretrained, fixed word vectors to create vectors for sentences, paragraphs and documents. Also known as paragraph2vec. Doesn't use a parse tree.
Implemented in gensim. See doc2vec tutorial

Deep Recursive Neural Networks for Compositionality in Language
Irsoy & Cardie. 2014. Uses Deep Recursive Neural Networks. Uses a parse tree.

Improved Semantic Representations From Tree-Structured Long Short-Term Memory Networks
Tai et al. 2015 Introduces Tree LSTM. Uses a parse tree.

Semi-supervised Sequence Learning
Dai, Le 2015 "With pretraining, we are able to train long short term memory recurrent networks up to a few hundred timesteps, thereby achieving strong performance in many text classification tasks, such as IMDB, DBpedia and 20 Newsgroups."

##Dialog A Neural Network Approach toContext-Sensitive Generation of Conversational Responses
Sordoni 2015. Generates responses to tweets.
Uses Recurrent Neural Network Language Model (RLM) architecture of (Mikolov et al., 2010). source code: RNNLM Toolkit

Neural Responding Machine for Short-Text Conversation
Shang et al. 2015 Uses Neural Responding Machine. Trained on Weibo dataset. Achieves one round conversations with 75% appropriate responses.

Towards AI-Complete Question Answering: A Set of Prerequisite Toy Tasks
Weston 2015. Classifies QA tasks. Builds on Memory Networks, and End-To-End Memory Networks which is implemented in MemNN

A Neural Conversation Model
Vinyals, Le 2015. Uses LSTM RNNs to generate conversational responses. Uses seq2seq framework (Sutskever, Vinyals, Le 2014). Used in Google Smart Reply feature for Inbox
seq2seq tutorial in TensorFlow