Michael Harris has spent the last decade immersed in open source technologies, big data, and architecture design for global brands. Upon joining Basis he has focused on real-world implementation of cutting-edge NLP technologies. As a child, he built a hang glider from plastic bags, duct tape, and sticks which subsequently disappeared the day before his planned flight from the roof of his parent’s garage, nevertheless, he has not lost his innovative spirit which he brings to every endeavor.
Michael Harris is speaking at the following session/s
Recognizing Names with AI using Solr Hashes and Stream Processing
Since the events of 9/11/2001, the US has successfully forestalled any major foreign terrorist attack of that magnitude. A key element of this success has been production and enforcement of national watchlists, such as for persons to be denied entry or passenger flight status. We will explore some of the techniques used in matching to those names, despite the enormous variation they undergo across languages, cultures, alphabetic scripts, and even deliberate obfuscation. This is a core case for AI, where accuracy is essential in traversing essentially human processes such as nicknaming, honorifics, and translation. It’s also deadly serious: false positives result in infringement to civil liberties, and a false negative resulted in mayhem in 2013 at the Boston Marathon. We will explore the means to process an essentially unlimited stream of text, exploring AI-driven high-recall and high-precision processes, to provide state-of-the-art accuracy in matching names to a watchlist.
The takeaway is understanding how to high-volume stream data in combination with Solr and AI for matching names in unstructured text with near real-time results. Technical attendees, developers, computational linguists, and intelligence organizations focused on name matching should be sure to attend.