Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/tomhosking/hercules
Hercules: Attributable and Scalable Opinion Summarization (ACL 2023)
https://github.com/tomhosking/hercules
nlp opinion-summarization summarization vq-vae
Last synced: 24 days ago
JSON representation
Hercules: Attributable and Scalable Opinion Summarization (ACL 2023)
- Host: GitHub
- URL: https://github.com/tomhosking/hercules
- Owner: tomhosking
- License: mit
- Created: 2022-12-02T15:13:22.000Z (almost 2 years ago)
- Default Branch: main
- Last Pushed: 2023-11-08T15:55:10.000Z (about 1 year ago)
- Last Synced: 2024-10-08T13:26:43.624Z (about 1 month ago)
- Topics: nlp, opinion-summarization, summarization, vq-vae
- Language: Python
- Homepage: http://tomho.sk/hercules
- Size: 81.4 MB
- Stars: 17
- Watchers: 1
- Forks: 3
- Open Issues: 2
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# [Hercules: Attributable and Scalable Opinion Summarization](https://aclanthology.org/2023.acl-long.473/)
[](https://paperswithcode.com/sota/unsupervised-opinion-summarization-on-space?p=attributable-and-scalable-opinion)
Code for the paper "[Attributable and Scalable Opinion Summarization](https://aclanthology.org/2023.acl-long.473/)", Tom Hosking, Hao Tang & Mirella Lapata (ACL 2023).
By representing sentences from reviews as paths through a discrete hierarchy, we can generate abstractive summaries that are informative, attributable and scale to hundreds of input reviews.
## Setup
Create a fresh environment:
```
conda create -n herculesenv python=3.9
conda activate herculesenv
```
or
```
python3 -m venv herculesenv
source herculesenv/bin/activate
```
Then install dependencies:
```
pip install -r requirements.txt
```
Download data/models:
- Space -> `./data/opagg/`
- AmaSum -> `./data/opagg/`
- [Trained checkpoints](http://tomho.sk/hercules/models/) -> `./models`
Tested with Python 3.9.
## Evaluation with trained models
See [`./examples/Space-Eval.ipynb`](examples/Space-Eval.ipynb)
or
```python
from torchseq.utils.model_loader import model_from_path
from torchseq.metric_hooks.hrq_agg import HRQAggregationMetricHook
model_slug = 'hercules_space' # Which model to load?
instance = model_from_path('./models/' + model_slug, output_path='./runs/', data_path='./data/', silent=True)
scores, res = HRQAggregationMetricHook.eval_generate_summaries_and_score(instance.config, instance, test=True)
print("Model {:}: Abstractive R2 = {:0.2f}, Extractive R2 = {:0.2f}".format(model_slug, scores['abstractive']['rouge2'], scores['extractive']['rouge2']))
```
## Training on SPACE/AmaSum from scratch
To train on SPACE, download the datasets (as above) then you should just be able to run:
```
torchseq --train --reload_after_train --validate --config ./configs/hercules_space.json
```
## Training on a new dataset (WIP)
You will need to:
- [ ] Install `allennlp==2.10.1` and `allennlp-models==2.10.1` via pip (ignore the warnings about version conflicts)
- [ ] Make a copy of your dataset in a format expected by the script below
- [ ] Run the dataset filtering scripts `./scripts/opagg_filter_space.py` and `./scripts/opagg_filter_space_eval.py`
- [ ] Run the script to generate training pairs `./scripts/generate_opagg_pairs.py`
- [ ] Make a copy of one of the training configs and update to point at your data
- [ ] Finally, train the model!
```
torchseq --train --reload_after_train --validate --config ./configs/{YOUR_CONFIG}.json
```
Please feel free to raise a Github issue or email me if you run into any difficulties!
## Citation
```
@inproceedings{hosking-etal-2023-attributable,
title = "Attributable and Scalable Opinion Summarization",
author = "Hosking, Tom and
Tang, Hao and
Lapata, Mirella",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.473",
pages = "8488--8505",
}
````