How Facebook’s patent identifying neologisms will change AI
Other than simply developing an automated urban dictionary, why is this significant? Because Facebook may be securing an strong advantage in its AI initiatives.
We’re seeing advancements in natural language processing through machine learning, but a system is only as good as its corpus and the ability to learn and adjust. Recommind established itself in the predictive coding space for legal review with its own patent in 2011, granting it “exclusive rights to workflow and technology for computer-assisted document analysis and review.” According to their website, “Predictive Coding uses machine learning to categorize and prioritize any document set faster, more accurately and more defensibly than contract attorneys, no matter how much data is involved.”
Facebook has its finger on the pulse of conversation across the globe, and can use that information to rapidly train and improve its systems. With this patent, it may be establishing itself as the best source of current language and usage, a key factor to strong conversational UI.
Whether Facebook chooses to make this ‘social glossary’ available to other developers as an open API (like IBM’s Watson), or keeps it proprietary remains to be seen. But as machine learning algorithms are open-sourced and data becomes the primary source of differentiation, Facebook is setting itself as a provider of exclusive — and potentially very valuable — content.