+The integration and evaluation of recommendations are based on the [STELLA evaluation framework](https://stella-project.org/). It allows for testing different recommender algorithms in real user environments. In GESIS Search, various recommender variants are integrated and evaluated against one another. We provide recommendations based on the similarity between documents, for example, the syntactic similarity between documents, such as titles and abstracts, or the semantic similarity when documents describe very similar content. In our evaluation scenario, the user decides on the best-performing recommender based on their click behavior. The more clicks a certain recommender achieves, the better it is evaluated.
0 commit comments