Automatic Extraction of Planetary Science Literature Authors, Targets and Compositions using Deep Dive and Tika
BACKGROUND:
There are continuous and growing contributions made to the body of knowledge representing discoveries and conclusions about various targets on the surface of Mars. This body of knowledge is represented by text, whether on the web in the form of HTML documents or as concise documents uploaded to the web. Planetary scientists and mission planners use search tools to update themselves about these new discoveries made about surface targets on Mars. However, these tools that mine the corpus of information are not currently built to meet the knowledge needs of scientists. Hence arises a need to build an information extraction system that can parse text and convert them into data usable by the mission planners and scientists. This data is the input that goes into building an encyclopedia for Mars targets.
This project is aimed at extracting the authors, targets and compositions of each contribution to the encyclopedia, which are essentially articles extracted from the web and/or other sources of data related to target discoveries on Mars. We will use Deep Dive integrated with Tika to extract this information and Deep Dive for reasoning based on the extracted data.
APPROACH:
Planetary science findings and conclusions exist in a vast body of knowledge in a number of formats. Apache Tika provides a generic API to detect and extract data from these multiple file formats.
As a first step in this project, we will use Tika to parse the target discovery data which are either in the PDF or HTML format. In the next stage, we will use Deep Dive to convert the parsed text from the literature into structured data which can be organized as discoveries and mentions of authors who have made contributions to literature related to such discoveries. This creates an automatic mapping from scholarly articles about target discoveries and conclusions on Mars to authors who have contributed to it.
We will thereby use Deep Dive hand in hand with Tika to parse text and create relationships between text and their corresponding authors.
STEPS TO INSTALL:
This is included in the Wiki for this project