【深度學習】情感分析和文本分類資料彙總

This document summarizes some potentially useful papers and code repositories on Sentiment analysis / document classification

• CNN:

  1. Related Paper: Convolutional Neural Networks for Sentence Classification. EMNLP 2014

     A Sensitivity Analysis of (and Practitioners' Guide to) Convolutional Neural Networks for Sentence Classification

     Code:

     Torch Code By harvardnlp

     tf CNN Seq Apply a 1-D convolutional network to classify sequence of words from IMDB sentiment dataset.

     keras CNN Seq Demonstrates the use of Convolution1D for text classification.

     tf Code

     Kefras code, Convolution with pretrained Glove embeddings Loads pre-trained word embeddings (GloVe embeddings) into a frozen Keras Embedding layer, and uses it to train a text classification model on the 20 Newsgroup dataset.

  2. Related Paper: Character-level Convolutional Networks for Text Classification NIPS 2015

     Code:

       Torch Code

  1. Related Paper: Molding CNNs for text: non-linear, non-consecutive convolutions. EMNLP 2015

     Code:

     Theano Code

  1. Related Paper: Very Deep Convolutional Networks for Text Classification. Facebook 2017      

• RNN:

  1. Related Paper: Semantic Compositionality through Recursive Matrix-Vector Spaces EMNLP2012

     Long Short Term Memory, Sepp Hochreiter & Jurgen Schmidhuber, Neural Computation 9(8): 1735-1780, 1997

     Code:

     Keras Recurrent Neural Network (LSTM) Trains a LSTM on the IMDB sentiment classification task.

     tf Recurrent Neural Network (LSTM) Apply an LSTM to IMDB sentiment dataset classification task.

     tf Dynamic RNN (LSTM) Apply a dynamic LSTM to classify variable length text from IMDB dataset.

  2. Related Paper: Text Classification Improved by Integrating Bidirectional LSTM with Two-dimensional Max Pooling COLING, 2016

     Code:

     Keras Bidirectional LSTM

     tf Bidirectional LSTM

  3. Related Paper: Semi-supervised sequence learning NIPS 2015

     Two approaches are presented by using unlabeled data pretraining to improve sequence learning with recurrent networks

  4. Related Paper: Document Modeling with Gated Recurrent Neural Network for Sentiment Classification EMNLP 2015

     Code:

     Java Code

  5. Related Paper: Effective LSTMs for Target-Dependent Sentiment Classification COLING 2016 Also Named as Target-Dependent Sentiment Classification with Long Short Term Memory

     Code:

     Java Code

  6. Related Paper: Adversarial Training Methods for Semi-Supervised Text Classification

• RCNN:

  1. Related Paper: Recurrent Convolutional Neural Networks for Text Classification. AAAI, 2015

     Code:

     Keras RCNN Trains a convolutional stack followed by a recurrent stack network on the IMDB sentiment classification task.

     Keras Code

• Others:

  1. Related Paper: Sequential Short-Text Classification with Recurrent and Convolutional Neural Networks NAACL 2016

  2. Related Paper: Bags of Tricks for Efficient Text Classification

     Code:

     keras example Trains a FastText model on the IMDB sentiment classification task.

     Facebook fastText FastText is a library for efficient learning of word representations and sentence classification

  3. Related Paper: Aspect Level Sentiment Classification with Deep Memory Network EMNLP 2016

     Code:

     Java Code