18. Beyond the Hype of Machine Learning and Artificial Intelligence Tue May 21, 2019, 1PM Scott Burt, President & CEO, Integro
Poll Question Do you think AI, ML, and/or NLP could help with InfoGov tasks?
UNIVERSAL ENHANCEMENTS INTEGRATED My theory on our use of AI for InfoGov GET EDUCATED ACCESSIBLE HANG ON BUT…
Common IG Use Cases Classification Retention Findability Value Cleanup Enrichment
WHAT IS AI? AI refers to machines that can learn, reason, and act for themselves
Is this AI? November 10, 2018, MIT Technology Review
WHAT IS MACHINE LEARNING? Looking for patterns in massive amounts of data
Document classification Machine Learning: Image recognition Looking for OCR patterns in massive Voice transcription amounts of data Document capture
Source: Box
Source: Box
WHAT IS NATURAL LANGUAGE Computer understanding, analysis and processing of human language PROCESSING (NLP)?
Source: https://azure.microsoft.com/en-us/services/cognitive-services/text-analytics/ See also: https://azure.microsoft.com/en-us/services/cognitive-services/language-understanding-intelligent-service/
SUPERVISED VS. UNSUPERVISED LEARNING
Unsupervised Learning Source: Brainspace
Supervised Learning Source: Integro Email Manager
IG USE CASES TO CONSIDER Seek repetitive tasks
Seek out low-skill tasks that occupy the time of high-skilled workers • Is the task data-driven? • Do you have the data to support the automation of the task? • Do you really need the scale that automation can provide? Kristian J. Hammond, Northwestern, Professor of Computer Science
Classification, Records and Defensible Disposal Source: Integro Email Manager
Classification Case Study | Email Archives F500 company • 1 Billion emails in archive • 87 records categories • 5 months to process the entire archive • 256M identified as records of 1B • Remainder emails not on Hold eligible for • defensible disposal
Classification Case Stud(ies) | Email Management with Human Oversight • Integro Email Manager™ Models trained to match full or partial file • plan Assists the user by auto classifying and • suggesting most likely categories Enables 3-zone email management with light • impact on users
Fortune 500 • pharmaceutical company • 40TB of classified content Training was ‘easy’ • considering the assembled corpus Processed and validated the • accuracy/quality of records categorization Classification Case Study | Records Audit
Text Capture for Enrichment
Privacy Compliance “Personal information” is defined under the CCPA as “information that identifies, relates to, describes, is capable of being associated with, or could reasonably be linked, directly or indirectly, with a particular consumer or household .”
Enrichment/Findability | Audio & Video Source: Box.com 26
THE LEADING Names you know. PROVIDERS
Natural Language Machine-Learning Artificial Processing (NLP) (ML) Intelligence (AI) LUIS (Language Azure Machine Azure AI, Text Understanding Learning Service, Analytics API Intelligent Service) Cubeflow, Cloud TPUs Amazon Amazon SageMaker Amazon Textract, Comprehend Amazon Rekognition Watson Natural Watson Machine Watson AI, Watson Language Learning, Watson Studio, Watson Understanding Visual Recognition Discovery Cloud Natural Cloud AI Platform AI, Language prepackaged solutions (e.g., Document Understanding AI, pretrained AI model for healthcare)
Pros and Cons Pros Cons Automate previously The time and effort to train • • of Using AI for impossible tasks – like the system InfoGov classify a billion documents Still early days • Cheap to start, just • Internal competing AI • subscribe projects No AI experience • Big vendors not focused on • ‘necessary’ IG Rapid progress! • Explicability & acceptance • AI as a feature in products challenges • – i.e., Box Skills, Integro Ethics and bias issues • Email Manager
Read… a lot • Look for opportunities – low risk, high • reward, bite-sized at first Get a free account with a cloud • provider Review current vendors and their • plans – The best first projects will likely be AI as features in products – Vendors will be seeking clients to be early adopters and references Engage product-neutral consulting • services Request a workshop and presentation • on Leveraging ML to Auto Classify Content for better Information Governance
Recommended ‘reading’
Poll Question Do you think AI, ML, and/or NLP could help with InfoGov tasks?
Forms data extraction • Text analysis • Text entity extraction • IG Use Content enrichment • Voice Recording transcription • Cases to Video transcription • consider Mergers and Acquisitions • Improving searchability and findability • Content Cleanup/Defensible Disposal • Email management •
Thank You Scott Burt President & CEO, Integro 720-904-1601 | sburt@integro.com Twitter: @integroburt LinkedIN: linkedin.com/in/scottburt
Recommend
More recommend
