Towards Extracting Detailed Metadata From Academic Research Articles

In this paper we present the results of a pilot project that addressed the problem of extracting relationships between publications, datasets, and methodologies. Our ultimate goal is to automatically extract details about specific datasets used, including variables and time periods as well as methodologies and techniques. The extracted metadata will help both researchers and organizations perform a number of important tasks, such as thorough and focused literature reviews, more efficiently and comprehensively. Furthermore, researchers will be able to generate new research applications and approach new datasets more easily. For organizations that host and generate datasets, the extracted metadata can provide an excellent starting point for evaluating dataset utilization at a detailed level.