samlukka / data2text-cbr

Data-to-Text Generation with Case-Based Reasoning

Data2Text-CBR

Data-to-Text Generation with Case-Based Reasoning

Requirements

Run:

pip install -r requirements.txt

How to Run

Download the GPT2 finetuned model

Download the fine-tuned GPT2 model from GDrive. It's a zip folder, unzip the files into a gpt2-finetuned folder in root directory.

Download the LaserTagger model

Download the trained LaserTagger model from GDrive. Put the contens of this zip folder into src/laserTagger/models folder.

You'll also need to download a pretrained BERT model from the official repository. You need to download the 12-layer ''BERT-Base, Cased'' model. Put the contents inside src/laserTagger/bert folder.

Note: there might be some issues with the TensorFlow version used in LaserTagger. You might need to run it in a virtua-environment then. Anyhow, even without LaserTagger generation can be done and there won't be any noticable difference in the metric scores.

Run

sh final.sh

1. Create clusters
2. Train Feature Weighting
3. Train important player classifier
4. Create Case-Base
5. Do generation
6. Apply LaserTagger

Cite

@inproceedings{upadhyay2021case,
  title={A Case-Based Approach to Data-to-Text Generation},
  author={Upadhyay, Ashish and Massie, Stewart and Singh, Ritwik Kumar and Gupta, Garima and Ojha, Muneendra},
  booktitle={International Conference on Case-Based Reasoning},
  pages={232--247},
  year={2021},
  organization={Springer}
}