Beyond Benchmarks: Assessing Knowledge Graph Completion Methods on Non-Benchmark Employee Data

This repository hosts the code and data for the paper titled "Beyond Benchmarks: Assessing Knowledge Graph Completion Methods on Non-Benchmark Employee Data," accepted at the International Conference on Data Science and Its Applications (ICoDSA) 2024. The research investigates the performance of Knowledge Graph Completion (KGC) methods—TransE, ComplEx, and TransH—on a non-benchmark employee dataset within the Human Resources (HR) domain. While existing studies primarily focus on benchmark datasets, this research aims to bridge the gap by evaluating KGC methods in real-world HR scenarios. By providing insights into the adaptability and efficacy of these methods beyond traditional benchmarks, this study contributes to a deeper understanding of KGC methods' performance within industry-specific applications.

Installation

To run the Python program associated with this repository, follow these steps:

1. Choose a directory on your local machine where you want to store the repository.

2. Clone this repository to your chosen directory using the following command:

   git clone https://github.com/khalidhabiburahman/kgc-non-benchmark-employee.git

3. Navigate to the cloned repository directory in your command prompt/terminal/local machine. For example, if you have cloned the repository into a folder named my-project, you can navigate to it using the cd command:

   cd path/to/my-project

4. Once you are in the repository directory, install the required Python libraries listed in the requirements.txt file using pip:

   pip install -r requirements.txt

This command will install all the necessary dependencies, including pandas, numpy, scikit-learn, matplotlib, rdflib, pathlib, and torch, ensuring that the program runs smoothly on your system.

Run the Experiments

To reproduce the experiments conducted in this research, follow the steps outlined below:

1. Raw Employee Data in Tabular Format

The raw employee data in tabular format serves as the initial input for the experiments. This file can be found in the folder Dataset/raw_tsel/diffbot-tsel-emp-export.csv.

2. Preprocess Raw Employee Data

Preprocessing of the raw employee data is performed using the raw_tsel_preprocess.ipynb notebook.

3. Construct Knowledge Graph

The construction of the knowledge graph involves the use of OntoText Refine. Download the tool and import the project file kg_tsel_emp.openrefine.tar.gz to observe the process of constructing the knowledge graph.

4. Preprocess KG Employee Data

Further preprocessing of the knowledge graph employee data is carried out using the kg_tsel_preprocess.ipynb notebook.

5. Training, Testing, and Validation Model

Training, testing, and validation of the model are conducted using the kgc_tsel.ipynb notebook. All these processes are encapsulated within a single notebook as the PyKEEN Pipeline is utilized for seamless execution.

By following these steps and executing the provided notebooks, you can replicate the experiments and analyze the results of the research.

Please note that this experiment utilizes Google Colab, which may require mounting to Google Drive for accessing the datasets. If this is not necessary for your setup, feel free to skip the code snippet related to mounting the drive.

Citation

If you use this package for published work, please cite the paper. Here is the BibTeX:

Yet to be published

License

This project is licensed under the Apache License 2.0 - see the LICENSE file for details.