Tensorflow implementation of attention-based LSTM models for sequence classification and sequence labeling.

Updates - 2017/07/29

• Updated code to work with the latest TensorFlow API: r1.2
• Code cleanup and formatting
• Note that this published code does not include the modeling of output label dependencies. One may add a loop function as in the rnn_decoder function in TensorFlow seq2seq.py example to feed emitted label embedding back to RNN state. Alternatively, sequence level optimization can be performed by adding a CRF layer on top of the RNN outputs.
• The dataset used in the paper can be found at: https://github.com/yvchen/JointSLU/tree/master/data. We used the training set in the original ATIS train/test split, which has 4978 training samples. There are 15 test samples that have multiple intent labels for an utterance. We used the more frequent label (most likely, "flight") as the true label during evaluation.

Setup

• TensorFlow, version r1.2 (https://www.tensorflow.org/api_docs/)

Usage:

data_dir=data/ATIS_samples
model_dir=model_tmp
max_sequence_length=50  # max length for train/valid/test sequence
task=joint  # available options: intent; tagging; joint
bidirectional_rnn=True  # available options: True; False
use_attention=True # available options: True; False

python run_multi-task_rnn.py --data_dir $data_dir \
      --train_dir   $model_dir\
      --max_sequence_length $max_sequence_length \
      --task $task \
      --bidirectional_rnn $bidirectional_rnn \
      --use_attention $use_attention

Reference

• Bing Liu, Ian Lane, "Attention-Based Recurrent Neural Network Models for Joint Intent Detection and Slot Filling", Interspeech, 2016 (PDF)

@inproceedings{Liu 2016,
author={Bing Liu and Ian Lane},
title={Attention-Based Recurrent Neural Network Models for Joint Intent Detection and Slot Filling},
year=2016,
booktitle={Interspeech 2016},
doi={10.21437/Interspeech.2016-1352},
url={http://dx.doi.org/10.21437/Interspeech.2016-1352},
pages={685--689}
}

Contact

Feel free to email [email protected] for any pertinent questions/bugs regarding the code.