Code for EMNLP 2020 Findings Paper: Language Generation via Combinatorial Constraint Satisfaction:A Tree Search Enhanced Monte-Carlo Approach

Requirement

• python

  • >=3.6

• python packages

  • PyTorch 1.2.0 or 1.3.0
  • numpy
  • pickle
  • Rake (pip install python-rake)
  • zpar (pip install python-zpar, download model file from https://github.com/frcchang/zpar/releases/download/v0.7.5/english-models.zip and extract it to POS/english-models)
  • skipthoughts (needed only when config.sim=='skipthoughts')
  • en (get en from https://www.nodebox.net/code/index.php/Linguistics and put under liguistics/)
  • Stanford CoreNLP (export CORENLP_HOME=/path-to-corenlp)
  • CoreNLP interface (pip install stanford-corenlp)
  • Huggingface Transformers (pip install transformers) 2.8.0
  • FastText 0.9.1 (cd fasttext; python setup.py install)

• word embedding

  • Download or train a word embedding first and place it at config.emb_path and set config.emb_path='word_max'.

Running

• Generation

  • For generating new sample for each tasks:
    • Give inputs in input/input.txt, or change the input parameter in config.py
    • Change the mode parameter in config.py: q for questions, i for imperative, n for sentiment;
    • Run python key_gen_search.py / sentiment_key_gen_search.py to generate.
    • Outputs are in output.

• Details

  • Make sure that paths for package and data are correctly set in 'config.py'.