We are excited to present OmniRec, a new open-source Python library for unified RecSys Experiments with popular libraries like Lenskit, RecBole, RecPack, and Elliot. ISG from the University of Siegen developed it together with our colleagues from GippLab at the University of Göttingen. Our team built OmniRec to address a persistent frustration in recommender systems research: the “reproducibility crisis” caused by fragmented data handling and inconsistent preprocessing across different libraries.
Currently, researchers often have to duplicate error-prone data preparation work, making it nearly impossible to fairly compare models across different frameworks.
OmniRec provides an all-in-one solution that standardizes these workflows, lowering the barrier to entry for newcomers and ensuring scientific rigor for experts.
Key Features of OmniRec
OmniRec acts as a central hub for the entire experimentation pipeline, from loading data to final evaluation. Its modular architecture is designed for transparency and ease of use.
- Access to 230+ Datasets: The library provides standardized, registered access to a vast library of datasets through a single interface.
- Write Once, Run Anywhere: Users define their preprocessing pipeline, including subsampling, filtering, and splitting strategies, just once. This pipeline then executes consistently across multiple integrated frameworks.
- Seamless Integration: The initial release features custom adapters for leading frameworks: RecPack, RecBole, Lenskit, and Elliot.
- No More Dependency Hell: OmniRec automatically manages isolated virtual environments for each library using the uv tool, preventing version conflicts between different research frameworks.
- Standardized Evaluation: A centralized Evaluator module ensures that metrics like nDCG, Recall, and RMSE are computed identically, regardless of which underlying framework trained the model
Proven in Research
OmniRec is the result of a multi-year development process within our lab. Its core logic has already served as the technical foundation for several of our papers published at premier venues such as ACM RecSys and ECIR.
Try it yourself:
https://omnirec.recommender-systems.com/
Paper: https://link.springer.com/chapter/10.1007/978-3-032-21321-1_18
Authors & Affiliations
- Lukas Wegmeth: University of Siegen, Germany
- Moritz Baumgart: University of Siegen and University of Göttingen, Germany
- Philipp Meister: University of Siegen and University of Göttingen, Germany
- Prof. Bela Gipp: University of Göttingen, Germany
- Prof. Joeran Beel: University of Siegen and Recommender-Systems.com, Germany
0 Comments