Skip to content

Latest commit

 

History

History
132 lines (93 loc) · 5.22 KB

README.md

File metadata and controls

132 lines (93 loc) · 5.22 KB
Raw

Neural Network Architectures For Open and Closed Literature-Based Discovery

In Natural Language Processing, Literature-based Discovery (LBD) is a form of knowledge extraction which aims to identify implicit relations by leveraging existing knowledge. Simply stated, LBD accomplishes the following task:

  • What:

    • Connects two pieces of knowledge previously thought unrelated.

  • How:

    • Identify implicit relationships between entities within disjoint texts.

Since it's inception in the 1980's by Dr. Don R. Swanson, many statistical methods to unearthing undiscovered relationships have been identified. These relationships include treatments for Parkinson’s Disease and multiple sclerosis, to understanding potential treatments for cancer, discovering new health benefits of curcumin, and treatments for migraine headaches, and identifying metabolites related to post-cardiac arrest. With the advent of deep learning, neural networks have achieved state of the art performance in various computer vision and NLP tasks. Our system Neural Network Architectures for Literature-based Discovery (NNLBD), integrates advances in deep learning for LBD.

Installation

NNLBD was developed and tested in Python version 3.6.x to 3.10.6. It also relies on the TensorFlow API. Versions 1.15.2 to 2.9.0 are supported. Tested operating systems to run our package include: Microsoft Windows 10 (64-bit) and Linux Mint (64-bit). The Microsoft Windows environment is not pre-package with Python. We recommend installing the appropriate Python version from python.org.

NOTE: Further mentions to Python refer to the Python3 installation environment.

Prior to instaling NNLBD, we recommend creating virtual environment.

  • Depending on how Python is installed in your system, one of the following commands will be appropriate:

    Linux:
        python -m venv <name_of_virtualenv>
        python3 -m venv <name_of_virtualenv>
Windows:
        python -m venv <name_of_virtualenv>

  • To verify which version of Python is installed, you can check via:

    python --version
python3 --version

Next, we activate your virtual environment and update your pip, setuptools, and wheel packages.

Linux:
       source <name_of_virtualenv>/bin/activate
       pip install -U pip setuptools wheel

Windows:
       "./<name_of_virtualenv>/Scripts/activate.bat"
       pip install -U pip setuptools wheel

NOTE: Powershell users will need to use the activate.ps1 script with suitable permissions or call cmd within powershell to execute the activate.bat script.

Python Requirements

After the setup of your virtual environment is complete, install the necessary NNLBD package requirements.

  • Python 3.10.x and TensorFlow 2.9.0 - (Recommended) 
    pip install -r requirements_mini_py3.10_tf2.9.0.txt
  • Python 3.6.x and TensorFlow 2.4.0 
    pip install -r requirements_mini_py3.6_tf2.4.0.txt
  • Python 3.6.x and TensorFlow 1.15.2 
    pip install -r requirements_mini_py3.6_tf1.15.2.txt

To manually install the required packages, execute the following commands:

  • Python v3.10.2 and TensorFlow v2.9.0 - (Recommended)

    pip install -U h5py==3.7.0 Keras==2.9.0 matplotlib==3.5.2 numpy==1.22.4 scipy==1.9.0 sparse==0.13.0 tensorflow==2.9.0

  • Python v3.6 and TensorFlow v2.4.0

    pip install -U h5py==2.10.0 Keras==2.4.3 matplotlib==3.3.4 numpy==1.19.5 scipy==1.5.4 tensorflow==2.4.0

  • Python v3.6 and TensorFlow v1.15.2

    pip install -U h5py==2.10.0 Keras==2.3.1 matplotlib==3.3.3 numpy==1.19.5 scipy==1.5.4 tensorflow==1.15.2 tensorflow-gpu==1.15.2

Getting Started, System Description and Model Details

To test your NNLBD environment, we recommend running one of our test scripts. We provide more information here.

To run one of our models, we provide details for getting started here. This also includes a description of our system, integrated models and a guide to replicate previous works.

We also provide an FAQ here.

Reference

@article{CUFFY2023104362,
   title = {Exploring a deep learning neural architecture for closed Literature-based discovery},
   journal = {Journal of Biomedical Informatics},
   volume = {143},
   pages = {104362},
   year = {2023},
   issn = {1532-0464},
   doi = {https://doi.org/10.1016/j.jbi.2023.104362},
   url = {https://www.sciencedirect.com/science/article/pii/S1532046423000837},
   author = {Clint Cuffy and Bridget T. McInnes}
}

License

This package is licensed under the GNU General Public License.

Authors

Current contributors: Clint Cuffy

Acknowledgments