Welcome 😄

This is the code used for generating the PS-ProofWiki. A dataset for Premise Selection in mathematical texts.

Paper: Deborah Ferreira, André Freitas, Natural Language Premise Selection: Finding Supporting Statements for Mathematical Texts, 12th Language Resources and Evaluation Conference (LREC), Marseille, France, 2020 (Language Resource Paper).

The dataset can be found in the folder dataset.

Run this command to download the data dump:

wget https://proofwiki.org/xmldump/latest.xml -O data/wiki.xml

Note: this might give you a slightly modified (updated) version of the dataset.

Run this command to install the requirements:

pip install -r requirements.txt

export PREFECT__FLOWS__CHECKPOINTING=true

After you installed all the requirements, run the extraction with

 python -m nlps_extraction.flows.extract_proofwiki

This will generate the files:

/output/train.json
/output/test.json
/output/dev.json
/output/kb.json

TF-IDF

python -m nlps_extraction.flows.baselines.run_tfidf_experiment

Doc2Vec

S-Bert

If you use our dataset or this pipeline to generate a new version of the dataset or use our dataset, please cite us:

@article{ferreira_freitas_lrec_2020,
 title={Natural Language Premise Selection: Finding Supporting Statements for Mathematical Texts.},
 journal={12th Language Resources and Evaluation Conference (LREC), Marseille, France.},
 author={Ferreira, Deborah and Freitas, Andre},
 year={2020},
 month={May}}

If you have any question, suggestions or ideas related to this dataset, please do not hesitate to contact me.

deborah[dot]ferreira[at]manchester[dot]ac[dot]uk