USPTO Grants H5 Patent for Its TAR Process
We interrupt your construction of a standard query designed to retrieve relevant documents, with high recall and precision, from a corpus of documents based on your expertise as an attorney and content expert in a case and your iterative work at refining a search query with reflexive search terms after your review of search results based on previous, iterative searching. Attention.
H5, a provider of e-discovery, technology-assisted review and case preparation support, has patented a process for "high recall and high precision relevancy searching." The U.S. Patent and Trademark Office granted to H5 on October 23 U.S. Patent No. 8,296,309, which covers aspects of H5's "deterministic" technology-assisted review, according to the company — and the USPTO. We checked.
The new business process patent covers H5's method for generating and refining relevance rules based on search queries to find relevant documents. According to H5 (and now the USPTO), the search and retrieval company now owns the process by which, after direction from counsel and informed by a manual review of example documents, rules are populated with linguistic search terms and iteratively refined over time and review of search results with an aim to increase accuracy and attain high recall and precision.
H5 believes its "deterministic" and "rules-based" TAR process is aligned with ZyLabs' rule-based TAR but distinguished from "predictive algorithmic machine-learning," according to an H5 fact sheet sent to LTN. The fact sheet states that H5's process to determine relevant documents is, among other things, based on "explicit rules" not algorithms that "infer relevance"; makes transparent not "black-box" coding decisions; and that H5 achieves high recall "and" high precision vis-á-vis predictive technologies that either get high recall "or" high precision.
H5's patent follows from its initial filing on May 28, 2010, which received the benefit of U.S. Provisional Patent Application No. 61/182,194, filed May 29, 2009, entitled "Human-Augmented Computer Cognition: User Modeling, Text Classification, and Their Optimization for High Recall with High Precision Information Retrieval."
Image courtesy of H5