Piranha is a text mining system developed for the United States Department of Energy (DOE) by Oak Ridge National Laboratory (ORNL). The software processes large volumes of unrelated free-text documents and shows relationships amongst them, a technique valuable across numerous scientific and data domains, from health care fraud to national security. The results are presented in clusters of prioritized relevance to business and government analysts. Piranha uses the term frequency/inverse corpus frequency term weighting method which provides strong parallel processing of textual information, thus the ability to analyze very large document sets. Piranha has six main strengths: Collecting and Extracting: Millions of documents from numerous sources such as databases and social media can be collected and text extracted from hundreds of file formats; This info. can then be translated to any number of languages. Storing and indexing: Documents in search servers, relational databases, etc. can be stored and indexed at will. Recommending: Recommending the most valuable information for particular users. Categorizing: Grouping items via supervised and semi-supervised machine learning methods and targeted search lists. Clustering: Similarity is used to create a hierarchical group of documents. Visualizing: Showing relationships among documents so that users can quickly recognize connections.
Contents
This work has resulted in eight issued ( 9,256,649, 8,825,710, 8,473,314, 7,937,389, 7,805,446, 7,693,9037, 7,315,858, 7,072,883), and several commercial licenses (including TextOre and Pro2Serve), a spin-off company with the inventors, Covenant Health, and Pro2Serve called VortexT Analytics, two R&D 100 Awards, and scores of peer reviewed research publications.