ConceptDrift

We introduce a novel framework, ConceptDrift, that models the hypothesis generation task as a sequence of temporal graphlets and simultaneously encodes spatial, temporal, and semantic change. Unlike existing approaches that treat these dimensions independently, ConceptDrift is the first to provide a holistic understanding of concept evolution by integrating them into a unified framework. Grounded in the theories of the Distributional Hypothesis and Conceptual Change, our method adapts these principles to the unique challenges of large-scale biomedical literature. We conduct extensive experiments across multiple datasets and demonstrate that ConceptDrift consistently outperforms state-of-the-art baselines in generating accurate and meaningful hypotheses. Our framework shows immediate practical benefits for web-based literature mining tools in life sciences and biomedicine, offering more robust and predictive feature representations.

• This repository holds the virology, neurology, and immunology datasets along with the implementation of ConceptDrift.

Data

Use Git-LFS to ensure data.zip is downloaded. Please unzip data.zip in the project directory. You should have a directory named data in the project directory.

Our temporal dynamic graphs are stored as Torch-Geometric TemporalData objects. You can find a pickle files holding the datasets along with mappings from the node ids to MeSH Terms in the data/{dataset} folders. The biobert embeddings for the terms are also provided in the data/{dataset} folders.

System

To ensure optimal performance and compatibility, your system should meet the following requirements:

• Python and Library Versions:
  • Python: Version 3.10 or higher
  • PyTorch: Version 2.0 or higher
  • Torch Geometric: Version 2.5.2 or higher
  • CUDA: Version 11.8
• Hardware Compatibility:
• ConceptDrift has been tested to successfully work under the following conditions:
  • CPU Systems: Single CPU with at least 32GB of RAM.
  • GPU Systems: Nvidia A40, A6000, or A100 GPU for accelerated computation.

Environment

To ensure ConceptDrift can run properly on your system, please follow the following steps:

1. To replicate our environment with all the necessary packages, please install the packages in requirements.txt in a virtual environment (e.g. Conda or Pip environment) before running our code.
2. Edit src/config.ini to include the file path to the data folder.

Training ConceptDrift

1. To train ConceptDrift, activate your environment with the necessary packages and go to the src folder.
2. Execute python train.py --dataset {dataset} to train on the virology, neurology, or neurology dataset.

Hyperparameters can be adjusted with the following command line arguments: --batch_size (default is 200), --max_epochs (default is 2), and --lr (default is 0.0001).