patent-parsing-tools 0.9.5

patent-parsing-tools is a library providing tools for generating training and test set from Google's USPTO data helpful with for testing machine learning algorithms

patent-parsing-tools

USPTO patents dataset generator.

Documentation

Read the docs

System requirements

sudo yum install python-devel libxslt-devel libxml2-devel

Installation:

pip install patent-parsing-tools

Examples:

Downloading dataset:

python -m patent_parsing_tools.downloader \
  --directory dataset \
  --year-from 2010 \
  --year-to 2010

Collecting and serializing data:

python -m patent_parsing_tools.supervisor \
  --working-directory patents/working_directory \
  --train-destination patents/train_destination \
  --test-destination patents/test_destination \
  --year-from 2014 \
  --year-to 2015

Generating dictionary with train set:

python -m patent_parsing_tools.bow.dictionary_maker \
  --train-directory patents/train_destination \
  --max-patents 1000000000 \
  --dictionary dictionary.txt \
  --dict-max-size 4096

Generate bag of words with train set and test set:

python -m patent_parsing_tools.bow.bag_of_words \
  --serialized-patents patents/train_destination \
  --destination-directory patents/final_dataset_train \
  --dictionary dictionary.txt \
  --batch-size 1048576
python -m patent_parsing_tools.bow.bag_of_words \
  --serialized-patents patents/test_destination \
  --destination-directory patents/final_dataset_test \
  --dictionary dictionary.txt \
  --batch-size 1048576

Testing

pytest

Contributing and develpment

$ mkvirtualenv ppt
$ workon ppt
(ppt) $ pip install -r requirements.txt

Publish new release

$ git tag v1.0
$ git push origin v1.0

Building documentation

(ppt) $ sphinx-build -M html docs docs_build

References

Usage:

• Elton, Using natural language processing techniques to extract information on the properties and functionalities of energetic materials from large text corpora, 2019, online: https://arxiv.org/abs/1903.00415.
• Lee, Natural Language Processing Techniques for Advancing Materials Discovery: A Short Review, 2023, online: https://doi.org/10.1007/s40684-023-00523-6.

License

The MIT License (MIT). Copyright (c) 2014 Michał Dul, Piotr Przetacznik, Krzysztof Strojny. Check LICENSE files for more information.