What? This web app recommends papers in several journals, together with US patents, given the DOIs or patent numbers of one or more items you like. It does this by finding similar articles, where similarity is based on the Euclidean distance between papers in an artificial 100-dimensional 'semantic space'. The position of a paper in this space is determined by the words and word combinations used in the abstract of the paper. The process of assigning these positions is known as latent semantic analysis. This implementation was inspired and assisted by Titipat Achakulvisut et al.'s 2016 work on Science Concierge.
Whence? The data was compiled with permission from SEG.org, and processed in Python. Many thanks to the SEG and especially to Isaac Farley for allowing me to use this dataset. The US Patent data is open data.
Why? The app is just for fun, and supports an article by Matt Hall in the March 2017 issue of The Leading Edge. It is part of a series of experiments hosted at geosci.ai.
When? The data was last updated on 27 Feb 2018.