CARET is a platform that is tailored to the development and evaluation of recommendation systems for software engineering. It provides Composable Analyses and Reusable Transformations to serve as a common ground for research on RSSE by providing reusable components that solve recurring tasks. The overall structure of the platform is sketched in the following Figure.
CARET is based on the enriched event streams (that are captured by FeedBaG) and simplified syntax trees, which cater for the specific needs of typical RSSE. CARET provides reusable components on top of SSTs that facilitate the construction of RSSE enabling researchers to focus their efforts on developing components that are specific to their recommendation problem at hand. It is designed to cater for the requirements of a wide range of source-based recommender systems that we have collected through a literature survey. CARET is not specific to a special kind of RSSE and its generic components allow the creation of future RSSE.
CARET offers the following three core features
- It can directly work with our published datasets.
- It features reusable components on top of SSTs that help researchers built RSSE, for example, generic analyses, transformations, and preprocessing steps.
- It provides support when building an evaluation pipeline for RSSE.