Exploration of Hugging Face Models by Heterogeneous Information Network and Linking Across Scholarly Repositories

Muhammad Asif Suryani, Saurav Karmakar, Brigitte Mathiak

Abstract: With the pervasive integration of Machine Learning (ML) focusing complex tasks across various domains, generally respective models and datasets are made available in numerous scientific repositories such as Hugging Face, GitHub for recognition and understanding within the research communities hence supporting open science initiative. However, the adaptability of these repositories is increasing among users hence raising a concern about the usability of these models and datasets effectively. Therefore, it is necessary to explore these repositories and compile comprehensive information that could facilitate users as well as repositories itself. Hugging Face, a leading repository, aims to furnish a platform that organizes and presents detailed information on models and datasets employed in research. As its adoption escalates within the research communities, the necessity to delve into such repositories becomes crucial to offer researchers valuable insights and promote efficient knowledge dissemination. This study focuses on exploring Hugging Face, particularly its machine learning models by exploiting various relevant features and the insights are presented in a Heterogeneous Information Network (HIN). Our research not only demonstrates the effectiveness of the available models on Hugging Face but also highlights potential links to relevant scholarly repositories by highlighting the significance of exploration which could contribute to the future integration of repositories, facilitating a more unified and accessible framework for scientific research.