Awesome Open Source
Awesome Open Source


This project is a pytorch implementation for CVAE_CGate model in my paper "Xu, Dusek, Konstas, Rieser. Better Conversations by Modeling, Filtering, and Optimizing for Coherence and Diversity", which sadly has been neither accpted by any conference nor put on the arxiv :(



Step1: Download the OpenSubtitles dataset

This code is based on OpenSubtitles dataset Automatic Turn Segmentation for Movie & TV Subtitles. To get the data, please contact the authors Pierre Lison. You should unzip the datset and name it as opensubtitles and put it in


Step2: Creat datasets for generator and discriminator

For the generator, a training pair consists of a dialogue context and a corresponding response. We consider three consecutive turns as the dialogue context and the following turn as the response. For the discriminator, positive examples are dialogue contexts with their following turn as the response, while negative examples are dialogue contexts with an utterance randomly sampled in the same dialogue as the response.

We use toolkit Opensubtitles processing tool owned by Ondrej Dusek to extract dialogues from OpenSubtitles dataset data/filter/opensubtitles/.

~/data/movie_tools/ -D -s -S train:train-dev:dev:test -r 97:1:1:1 -d all_dialogues_cased opensubtitles/ dial.jsons.txt

The outputs are

  • train.dial.jsons.txt
  • train-dev.dial.jsons.txt
  • dev.dial.jsons.txt
  • test.dial.jsons.txt

as the split ratio 97:1:1:1 with format of one dialogue per line

["utterance 1", "utterance 2", "utterance 3"...]

for example,

["Watch out !", "Oh , what fun !", "JON :", "That was fun .", "Oh , that was great !", "Oh , time for a break ?", "Dad , I 'm hungry .", "I 'm really hungry .", "Can we eat now ?", "Keep your shirt on .", "We 'll be in Potter 's Cove in 20 minutes .", "OK , how about some pictures ?", "Here we go .", "Everybody smile .", "Say cheese ."]

Then, we construct the training dataset for generator and discriminator from train.dial.jsons.txt by running


The outputs are

  • train.en inputs of encoder in generator (dialogue contexts)
  • outputs of decoder in generator (expacted responses)
  • train.pos positive examples for discriminator (dialogue contexts with their following turn as the response)
  • train.neg negative examples for discriminator (dialogue contexts with an utterance randomly sampled in the same dialogue as the response)

The format of train.en is utterance1 <u2> utterance2 <u1> utterance3 in each line, for example

well , i 'm glad you called me . <u2> i 'm not . <u1> no , you did the right thing .

The format of is response in each line, for example

you 'll protect him , won 't you ?

The formats for train.pos and train.neg are the same utterance1 <u2> utterance2 <u1> utterance3 \t response, for example

pull up sooner . <u2> ok , skipper ! <u1> do you think they 'll ever get it ?	     give them a week .

At last, we randomly sample 5000 cases for train-dev, dev, test separately by running following commands and outputs for each set are similar with training set.

  • python train-dev for train-dev set
  • python dev for dev set
  • python test for testing set

Step3: Filter the training set for generator

Step3.1: Train GloVe model on OpenSubtitles

Run the following command in data/filter/ to read subtitles from json files and save in file bag_of_words in the same directory.


Then, run the following two commands to train a GloVe model on the OpenSubtitles dataset. is used to build the corpus model corpus.model and train the model on corpus.model. The trained model is glove.model in the same directory.


Step3.2: Filter the training set for generator

python train

The outputs for this command is cosine distance of the two semantic vectors of a dialogue context and its response (Eq.1 in the paper). The format is cosine distance \t dialogue context \t response. For example

0.9228650507713863	  they tell the whole story . <u2> i sent them , but i want the weekend . <u1> please , mr president .	    only at the weekend .

Then you can filter training pairs with lower coherence score (cosine distance) and rewrite the train.en file with the filtered dialogue contexts and file with their responses.

Step4: Training for Generator (and Discriminator)

Step4.1: Data copying

You need to copy the following data from data/filter/ to data/.

  • train.en
  • dev.en
  • test.en

Step4.2: Dictionary building and data preparation


This command will create three socuments in data/.


Step4.3: Discriminator training

cd get_c/

This command will create glove.model in get_c/.

Step4.4: Generator training

Now, go back to the main directory. Run the following command to train the CVAEf_CGate generator

python -data data/dialogue -save_model dialogue-model -epochs 30 -report_every 100 -batch_size 128 -dropout 0.2 -src_word_vec_size 128 -tgt_word_vec_size 128 -rnn_size 128 -global_attention general -input_feed 0 -glove_dir get_c/glove.model -learning_rate 1 -context_gate both

The trained models are named as dialogue-model_acc_*

Step5: Inference

python -model $MODEL -src data/test.en -tgt data/ -report_bleu -verbose

The predictions are saved in file pred.txt

Get A Weekly Email With Trending Projects For These Topics
No Spam. Unsubscribe easily at any time.
Python (1,140,932
Deep Learning (23,665
Pytorch (11,586
Nlp (8,360
Chatbot (3,505
Deep Neural Networks (2,860
Nlp Machine Learning (1,221
Seq2seq (544
Variational Autoencoder (372
Sequence To Sequence (188
Dialogue (163
Dialogue Generation (45
Emnlp (36
Emnlp2018 (17
Related Projects