Since recommender systems play an important role in our online experience today and are involved in a wide range of decisions, multiple stakeholders are requesting explanations for the corresponding algorithmic predictions. These demands—together with the benefits of explanations (e.g., trust, efficiency, and sometimes even persuasion)—have triggered significant interest from researchers in academia and in industry. Nonetheless, to the best of our knowledge, no comprehensive toolkit for development and evaluation of explainable recommender systems is available to the community yet. Instead, researchers are frequently faced with the challenge of re-implementing prior algorithms when creating and evaluating new approaches. This paper introduces recoXplainer, an easy-to-use, unified and extendable library that supports the development and evaluation of explainable recommender systems. recoXplainer includes several state-of-the-art black box algorithms, model-based and post-hoc explainability techniques, as well as offline evaluation metrics in order to assess the quality of the explanation algorithms.

中文翻译：

---

RecoXplainer：用于可解释推荐系统的开发和离线评估的库


由于推荐系统在我们当今的在线体验中发挥着重要作用，并且涉及广泛的决策，因此多个利益相关者要求对相应的算法预测做出解释。这些要求以及解释的好处（例如信任、效率，有时甚至是说服力）引起了学术界和工业界研究人员的极大兴趣。尽管如此，据我们所知，社区尚未提供用于开发和评估可解释推荐系统的综合工具包。相反，研究人员在创建和评估新方法时经常面临重新实现现有算法的挑战。本文介绍了 recoXplainer，这是一个易于使用、统一且可扩展​​的库，支持可解释推荐系统的开发和评估。 recoXplainer 包括多种最先进的黑盒算法、基于模型和事后可解释性技术，以及离线评估指标，以评估解释算法的质量。