Deep Insights - Search Analytics for the Domain Sciences


Much of the data in within the domain sciences is textual in nature. Let that sink in. It's certainly something that with the wealth of remotely sensed, airborne and spaceborne, and in-situ scientific measurements that a casual user wouldn't expect, but it's precisely the case.

Take the polar data repository Advanced Cooperative Arctic Data and Information Service (ACADIS) as a prime example - 50% of its data is text-based and not stored in scientific binary data files.

The thing about text - there's a lot of meaning that can derived from automatically extracting and processing text. Persons, names, organizations, locations, dates, all of the types of features that are extremely useful in answering questions such as What are the impacts of Atmospheric Carbon per year as the relate to the Polar Sciences or show a trend of mentions of Oil Spills in the Arctic Regions?

Through funding from the NSF, DARPA and NASA we have created a set of Polar Deep Insights that are information-retrieval and machine learning-based extractions from textual and scientific information in the Polar and Cryospheric sciences domain that answer grand challenge questions such as those above.The extracted information is made available and interactive using the Data-Driven Documents (D3) framework and in our Polar Deep Insights we demonstrate that we can answer grand challenge questions from the President's Strategy for the Arctic Region examples of which were previously mentioned above.

 In this session we will demonstrate our system, and invite the community to contribute to prioritizing what types of questions we should be focusing on.  


Mattmann, C.; Deep Insights - Search Analytics for the Domain Sciences; 2016 ESIP Summer Meeting. ESIP Commons , March 2016