Research and experimentation in various scientific fields are based on the observation, analysis and benchmarking on datasets. The advancement of research and development has thus, strengthened the importance of dataset access. However, without enough knowledge of relevant datasets, researchers usually have to go through a process which we term \manual dataset retrieval". With the accelerated rate of scholarly publications, manually finding the relevant dataset for a given research area based on its usage or popularity is increasingly becoming more and more difficult and tedious. In this paper, we present Delve, a web-based dataset retrieval and document analysis system. Unlike traditional academic search engines and dataset repositories, Delve is dataset driven and provides a medium for dataset retrieval based on the suitability or usage in a given field. It also visualizes dataset and document citation relationship, and enables users to analyze a scientific document by uploading its full PDF. In this paper, we first discuss the reasons why the scientific community needs a system like Delve. We then proceed to introduce its internal design and explain how Delve works and how it is beneficial to researchers of all levels.