Developing Custom-Made Comment-Recommendation Prototypes with a Modular Design Framework

Comment sections of news articles are a popular way to discuss the contents of these articles. But the number of comments posted every day has become so large that almost no one can get a solid overview about the discussion. To address this problem, there are many approaches for comment recommendation systems. However, they tend to focus mostly on the development of sophisticated models to combat this problem while evaluating their systems in limited and mostly artificial settings. In our paper, we introduce a modular open-source software framework for the development of comment recommendation prototypes that can be used to evaluate models in real-world environments. The modularity allows developing systems that are adapted exactly to the use-case or model one needs. This concept allows exchanging and adapting the different components of the concept to test e.g. different user-interfaces or recommendation models. To show the usability of our framework we present the implementations of two comment-recommendation applications.