algoprog / faspect

A library for open domain query facet extraction and generation

Faspect

Faspect is a library with various model implementations for open domain query facet extraction and generation. For more details you can have a look at our ICTIR 2022 paper: Revisiting Open Domain Query Facet Extraction and Generation.

Installation

pip install -r requirements.txt

Install the required nltk packages

python -m nltk.downloader stopwords averaged_perceptron_tagger

You can also try an online demo on Colab before running locally.

Usage

Run faspect.py and use the API to extract facets, for example you can post json data in this format to 127.0.0.1:6000/extract:

{
    "query": "cars",
    "documents": ["doc 1 text", "doc 2 text", "doc N text"]
}

and get results:

{
    "facets": [
       "cars for rent",
       "animated movie",
       "cars for sale"
    ]
}

or use the classes from the models folder in your project. For evaluation of your own models, please use the metric implementations in evaluation.py.

Citation

If you use any part of this code, including model weights, please cite our paper:

@article{Samarinas_2022_Faspect,
  title   =  {Revisiting Open Domain Query Facet Extraction and Generation},
  author  =  {Chris Samarinas and Arkin Dharawat and Hamed Zamani},
  journal =  {Proceedings of the 2022 ACM SIGIR International Conference on Theory of Information Retrieval},
  year    =  {2022}
}

For any problems or suggestions, feel free to open an issue or contact Chris Samarinas.

faspect's Issues

Bert_score issue

There will be something wrong when i calculate bert_score if len(cand_terms) != len(gt_terms)

I got higher results

Why i got higher results using your uploaded checkpoint and code.
Here are my results: