About 2012, I did a literature research on the following topic but unfortunately lost my results. Specifically one paper comes repeatedly to my mind, therefore maybe someone knows about this or subsequent research.
The research question of that paper was about information retrieval, specifically full text mining of scientific literature to automate literature review.
The researchers built a custom spider/search engine which takes as input not a set of keywords but set of documents.
They described that this approach allowed them to find relevant branches of research they were not aware of.
Which paper was it and/or what is state of the art regarding this topic?