Latent-Anchor-Plan

Code and Data for our Findings of EMNLP 2020 paper titled 'Narrative Text Generation with a Latent Discrete Plan'

Code

Added to code/

Relevant scripts can be found at code/scripts/

Data

Processed data can be found at Link.

Copy the data from the above link to data/ folder

Processed vocab file: Link -> saved_model_vocab_file -> vocabs/vocab.pkl

Saved Model

We also share trained model file Link -> saved_model_vocab_file -> models/

Download the model file, and move to to code/tmp/models/ location.

Run the sampling and evaluation scripts at code/scripts/lap.sh

Requirements

• python 3.7.2
• pytorch 0.4.1.post2

Reference

@inproceedings{jhamtani2020latentplan, 
title={Learning to Explain: Datasets and Models for Identifying Valid Reasoning Chains in Multihop Question-Answering}, 
author={Jhamtani, Harsh and Berg-Kirkpatrick, Taylor}, 
booktitle={Findings of EMNLP 2020}, 
year={2020} 
}

Our code and data is based on work of Yao et al 2019. If you use the code or processed data, also consider citing :

@inproceedings{yao2019plan,
  title={Plan-and-write: Towards better automatic storytelling},
  author={Yao, Lili and Peng, Nanyun and Weischedel, Ralph and Knight, Kevin and Zhao, Dongyan and Yan, Rui},
  booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
  year={2019}
}