Training Tagger and Generator

• This repository contains the code to train the tagger and generator modules.

• Apart from scripts to train the modules, it also has scripts needed to run inference on the test set and to run evaluation for metrics like BLEU, ROUGE, and METEOR.

---

Background

• Both tagger and generator are seq2seq models that require parallel data generated by the data prep module.

• The parallel datasets are:

  • Tagger: entagged_parallel.{split}.en → entagged_parallel.{split}.tagged
  • Generated: engenerated_parallel.{split}.en → engenerated_parallel.{split}.generated (where {split} is either train, test, or dev.)

---

Prepare BPE

bash scripts/prepare_bpe.sh [tagged|generated] {base_folder}

Where:

• base_folder: The folder in which the data files are stored (argument used in creation of training data)

---

Train Tagger and Generator

bash scripts/train_tagger.sh tagged {handle} {base_folder}

Where:

• handle: This is an identifier used to bucketize models trained on different datasets. Models on each handle are stored seperate folders with names indexed by {handle}, within the {models} directory.
• base_folder: The folder in which the data files are stored (argument used in creation of training data).

Train Generator

bash scripts/train_generator.sh generated {handle} {base_folder}

Where:

• handle: This is an identifier used to bucketize models trained on different datasets. Models on each handle are stored seperate folders with names indexed by {handle}, within the {models} directory.
• base_folder: The folder in which the data files are stored (argument used in creation of training data).

---

Inference

bash scripts/inference.sh {input_file} {jobname}\
                          tagged generated\
                          {handle}\
                          {style_0_label} {style_1_label}\
                          {base_folder} {device}

Where:

• input_file: The input test file which needs to be transferred. This is the raw text file, with one sentence per line.
• jobname: A unique identifier for the inference job.
• handle: dataset argument we pass when we train tagger or generator -- used to identify model paths for tagger and generator.
• style_0_label: A label for style 0
• style_1_label: A label for style 1
• base_folder: The folder in which the data files are stored (argument used in creation of training data)
• device: gpu id

---

Evaluation

bash run_context_eval.sh {hypothesis_filepath} {reference_filepath}

Where:

• hypothesis_filepath: The full path to the transferred output from trained model (hypothesis).
• reference_filepath: The full path to the ideal output (for BLEU-r) or the original input file (for BLEU-s).

Trained Models

The trained models can be found here.

References

• The code for evaluation has been partially borrowed from https://github.com/Maluuba/nlg-eval
• Most of the code for the training pipeline has been borrowed from https://github.com/pmichel31415/jsalt-2019-mt-tutorial