Scientific Discovery as Link Prediction in Influence and Citation Graphs

Case ID:

With the wealth of scientific information available online, understanding where the gaps are in research is crucial. This invention looks to identify those white spaces within specific scientific fields. Under the umbrella of natural language processing, this invention can analyze texts and help to predict from where scientific expansions will come. The key attribute of the invention is that it utilizes two graphs; one of these graphs is for citations, coming from readily available scientific literature and the other for influence. The influence graph is then extracted and the resulting influence graph to test predictions using back testing. The invention has a wide range of uses including in academia to help scientists to identify where they should focus research efforts and for companies looking to expand in a particular field. A key advantage is the ability to accurately predict otherwise hidden and important scientific discoveries in topics that are not sufficiently studied. For example, out of the top ten 2012 predictions derived from this invention, 80% proved correct. 



The amount of published papers in any given field that can be found online is astounding, such as the 25 million biomedical publications that can be found in the PubMed database. Although vast amounts of published research is available thanks to the internet and its sharing culture, utilizing this research to predict where the science is heading is challenging for companies and researchers. This is where this innovation comes into play. Future discoveries can now be more accurately predicted which can help researchers and companies plan the future. In fact, it can help them become the future.   



  • Research portfolio management
  • Aiding faculty research to show what spaces of their field need to be developed
  • Companies looking to expand their businesses within their current fields by finding the white spaces of that field



  • Provides an opportunity to invest in near term white space identified by this predictive tool
  • Ability to identify insufficiently studied topics
Patent Information:
Contact For More Information:
Jay Martin
Licensing Associate, Software and Copyright
The University of Arizona
Lead Inventor(s):
Mihai Surdeanu
Fan Luo
Marco Antonio Valenzuela Escarcega
Gustave Hahn-Powell