The prolific inventors at Unified Compliance (UC) have been granted a new patent on Multiword Expression (MWE) detection, the 13th of their patents, totaling 257 issued claims to date.
LAS VEGAS, Aug. 2, 2022 /PRNewswire-PRWeb/ -- The prolific inventors at Unified Compliance (UC) have been granted a new patent on Multiword Expression (MWE) detection, the 13th of their patents, totaling 257 issued claims to date. The new patent detects multiple-word noun phrases and multiple-word verb phrases. The UC team leverages this patented technology in their Application Programming Interfaces (APIs) to automatically identify multiword expressions in various Governance, Risk, and Compliance (GRC) content they examine (Citations, Controls, Configuration Guidance, Roles, etc.).
This patented technology and methodology can identify a more significant number of MWEs than previous Natural Language Processing (NLP) dictionaries and systems because it employs multiple models employing different approaches for identifying MWEs. This effectively puts five or more sets of eyes on each Citation being examined. In addition, some models have their own internal voting systems for making decisions about MWE detection. Using this multiple model approach, the MWE APIs can identify a much more complete list of noun phrases and/or verb phrases. One model even has the capability to flag sections that it deems to need a human touch. "By applying this combination of artificial intelligence and augmented intelligence, our team at UC can expand our current mapping library more efficiently and accurately," says Dorian Cougias, CEO. The accuracy of the new patented process using the combined approach (transformer, noun, and verb chunk linguistic algorithms) increases the reliability from 40.76% to 73.52%.
With this division of labor between the multiple models and expanding the current library, the new MWE detection is also more efficient than previous methods because human interactions needed to intervene when the system "gets it wrong" are reduced by 30%. This allows for faster mapping of new authority documents that are added to the CCH library.
Additionally, this new approach auto-suggests new terms for the federated Compliance Dictionary. When the MWEs detected by the API aren't found in the dictionary, the API automatically suggests they be added and provides the context in which the term was found. This takes the guesswork out of deciding when to add multiword expressions to a dictionary.
UC will be rolling out a new API gateway featuring their newly patented MWE detection technology beginning in August of this year. Further specifics about the new patent "Automatically identifying multiword expressions" can be found at the following web address. https://patents.google.com/patent/US11386270B2/en?oq=11386270 The UC team has also released a peer-reviewed scientific paper on Multiword Expression identification that can be found at https://www.researchgate.net/publication/343162894
About Unified Compliance
Unified Compliance (UC) provides the world's most vetted compliance framework and supporting APIs to the Global 2000. Their Unified Compliance Framework® (UCF®) is the world's most extensive library of over 1000 interdependent regulatory compliance documents and the world's only commercially accessible Common Controls framework. AWS, Verizon, Google, JPMorgan, NASA, US Treasury, and hundreds of other organizations across industries rely on UC's multi-patented processes and application programming interface (API). The UCF incorporates artificial and augmented intelligence to simplify, strengthen, and support compliance processes, ensuring greater ease of attestation and compliance success. Unified Compliance's logical, scientific approach to compliance simplifies the process. It helps compliance professionals gather the evidence they need to prove compliance in the most cost-efficient method available on the market. Detailed information is at https://www.unifiedcompliance.com
Jody Mack, Moxy LLC, 1 (702) 776-9898, [email protected]
SOURCE Unified Compliance (UC)