Program Chair: Tom Reamy Chief Knowledge Architect Text Analytics Forum 2018 Program Chair: Tom Reamy Chief Knowledge Architect KAPS Group http://www.kapsgroup.com Author: Deep Text
Text Analytics Forum (TAF) Introduction Conference Highlights and Themes Second year – following an inaugural success KMWorld – TA is a means of enriching KM Enriched content, expertise, collaboration TBC – TA Minds the Gap between taxonomy and content New knowledge organizations, cognitive-based ESD – TA is best means of improving search Faceted search, semi-automated subject tagging SharePoint/ O365 – all major TA vendors integrate with it Hybrid model – software characterizing document, sent to author/editor for human check
Text Analytics Forum (TAF) Introduction Conference Highlights and Themes Overview of field of text analytics Keynote Panel on Cognitive Computing Two tracks – technical / applications Technical ML and Rules, TA and Cognitive Computing, New Techniques, Auto-tagging and Entity Extraction, Modeling Tacit Knowledge Business / Applications TA Basics, Business Case, Practical AI, Structure in Documents, Q&A systems, Multiple Case Studies – Space to Human Trafficking Ask the Experts Panel Some questions about the field of TA We want your questions Book Signing TU – 3:30 – 4:15, Wed – 3:15-4:00
Agenda Text Analytics Introduction What is it? What is it good for? Key Ideas – Present and Future Content Structure Deep Learning and Symbolic AI Obstacles in Text Analytics Adoption Selling the Benefits – new approach Questions
Text Analytics Forum (TAF) Introduction Deep Text: The Book – Who Am I? KAPS Group – 13 years, Network of consultants (“hiring”) Taxonomy to text analytics Consulting, development – platform and applications Strategy, Smart Start, Search, Smart Social Media TA Training (1 day to 1 month), TA Audit Partners – Synaptica, SAS, IBM, Expert System, Smartlogic, etc. Clients: Genentech, Novartis, Northwestern Mutual Life, Financial Times, Hyatt, Home Depot, Harvard, British Parliament, Battelle, Amdocs, FDA, GAO, World Bank, Dept. of Transportation, IMF, etc. Presentations, Articles, White Papers – www.kapsgroup.com
A treasure trove of technical detail, likely to become a definitive source on text analytics – Kirkus Reviews
Text Analytics Forum (TAF) Introduction What is Text Analytics? Text analytics is the use of software and knowledge models to analyze/utilize poly-structured text. Text Mining – NLP, statistical, predictive, machine learning Different skills, mind set, Math & data not language Annotation/Extraction – entities and facts – known and unknown, concepts, events - catalogs with variants, rule based Sentiment Analysis Entities and sentiment words – statistics & rules Summarization Dynamic – based on a search query term Document – based on primary topics, position in document
Text Analytics Forum (TAF) Introduction What is Text Analytics? Auto-categorization = the brains of the outfit Training sets – Bayesian, Vector space Terms – literal strings, stemming, dictionary of related terms Boolean– Full search syntax – AND, OR, NOT Advanced – DIST(#), ORDDIST#, PARAGRAPH, SENTENCE Applications: Search, BI, CI, Financial Services, eDiscovery, etc. Fraud – Function word patterns / Fake News Sentiment analysis – beyond positive and negative Whole new applications – customers likely to cancel, new? Platform for multiple features – Sentiment, Extraction Disambiguation - Identification of objects, events, context Fact Extraction – context around words, concepts
Text Analytics Forum (TAF) Introduction Trends Market - 3.5 B to 10.5B 2023 Cloud Technology Growing Real time analytics, text from anywhere BOTs in the enterprise Conversational Interfaces Social – Behavioral Analytics New models of people New Knowledge Organizations K Graphs Emotion, Motivation Taxonomies AI and ML – hype? Or ?
Text Analytics Forum (TAF) Introduction Key Ideas / Trends in Text Analytics
Text Analytics Forum (TAF) Introduction Key Ideas in Text Analytics Poly-structured text – Content Types and Sections Deep Text a foundation for multiple applications Using sections for better auto rules AI / Machine Learning AND Rules Obstacles for Text Analytics New Approach to Selling the Benefits of Text Analytics
Text Analytics Forum (TAF) Introduction Adding Structure to Unstructured Content Content Type – defined by sections Blogs, Announcements, Articles, Press Releases, News, Case Reports, Correspondence Sections Metadata and text indicators – rules to find Document Level: Title-Keywords, Abstract, summary, etc. Special sections – Methods, Objectives, Results, etc. Data patterns – dates, addresses – need context rules Weights – ignore all but section text to sophisticated weighting Clusters and machine learning – at section level, not document Clusters as sections, clusters within sections Worst method – strip out all indicators and use a Bag of Words
Text Analytics Forum (TAF) Introduction Key Ideas in Text Analytics Poly-structured text – Content Types and Sections AI / Machine Learning AND Rules Data AND Concepts Obstacles for Text Analytics New Approach to Selling the Benefits of Text Analytics
Text Analytics Forum (TAF) Introduction Key Ideas in Text Analytics: Deep Learning & ML Neural Networks – from 1980’s New = size and speed, larger networks = can learn better and faster Machine Learning – Deep Learning but less and more Limited granularity – high level categories, very orthogonal Faster to get started and get to 60% - then the wall Biggest advantage = ML can learn – connection strength Strongest in areas like image recognition, physical patterns Weakest – concepts, subjects, deep language, metaphors, etc. Deep Learning is a Dead End - accuracy – 60-70% - concepts
Text Analytics Forum (TAF) Introduction Deep Text vs. Deep Learning Black Box – Watson – “We don’t know how or why it works” Susceptible to bias – hard to fix Domain Specific, data not deep understanding No common sense & no strategy to get there (faster not enough) Do rules take more effort to develop? Some studies show it is less Rules-Based Rule-based system reported 92 percent accuracy and a fourfold increase in productivity. Less up front cost, and less time spent refining Transparent – can fix, refine
Text Analytics Forum (TAF) Introduction Combine ML and Rules Future = Combine machine learning and rules Application Level to categorization language level Cognitive Model – Thinking Fast and Slow Automatic answers – Good until they are not Slow – complex problem solving Add content structure/ contexts – makes both smarter Rules – right kind of rules stable logic, variable text IF pattern > Threshold THEN X Hierarchical Networks Rules that learn
Text Analytics Forum (TAF) Introduction Key Ideas in Text Analytics Poly-structured text – Content Types and Sections AI / Machine Learning AND Rules Obstacles for Text Analytics Survey Says New Approach to Selling the Benefits of Text Analytics
Text Analytics Forum (TAF) Introduction Obstacles to Text Analytics Survey Says: Lack of Knowledge & Standards Text is an order of magnitude more complex than data Entities – Facts - Concepts No SQL for text Python is lower level language Common functions, disparate syntax Requires interdisciplinary – programming and language skills Rare that organizations have this collaboration set up No certification for TA development Vendor specific training Software is expensive and difficult
Text Analytics Forum (TAF) Introduction Obstacles to Text Analytics: Possible Solutions Buy my book! Develop and adopt standards Simplify development process but keep power Develop better training Not just one week of software training Better types of rules Theoretical Breakthrough – ML AND Rules Major Power takes over – Adobe, Microsoft, ? Hybrid solutions – Apple and Goggle – 1,000’s of humans Good, But still a hard sell
Text Analytics Forum (TAF) Introduction Key Ideas in Text Analytics Poly-structured text – Content Types and Sections AI / Machine Learning AND Rules Obstacles for Text Analytics New Approach to Selling the Benefits of Text Analytics Seeing is Believing
Benefits of Text Analytics Selling the Benefits - Overview Start with numerical studies - ROI Stories – Pharma example Stories – find own real life stories Focus of Stories: Business Objectives Selling to C Level Different language Need to educate – what it is and why Internal Advocacy is Key BUT – Fatal Flaw – Hard to believe!
Text Analytics Forum (TAF) Introduction Selling the Vision New Approach – Mini-POC One approach – One week Elements Taxonomy (Old, one branch) – 10-20 nodes to 100 Sample content – 10-20 documents per node Simple content model – document sections Build categorization rules for all nodes Cat rules faster, easier, better based on sections Demo – Simple search (15%-50%) to 90%+ (Story)
Text Analytics Forum (TAF) Introduction Benefits: Selling the Vision: Mini-POC Something that people can see, touch, play with Real application with real content See the value of Taxonomy + Text Analytics Appeal to all audiences – Librarians to KM to technology geeks to executives Option – Comparison with fully automatic clusters Start of building a foundation for full enterprise Full POC can build (most of) that foundation
Text Analytics Forum (TAF) Introduction Conclusions No such thing as unstructured text Content Models / Sections capture and utilize structure Power of AND Still early – ML and rules Wonderful mix of hype and reality AI-Deep Learning – still “Two years away” for concepts Data is here and getting smarter Main obstacles are complexity of language and lack of knowledge New Approach to educate and get buy in – Mini-POC Enjoy the conference!
Questions? Tom Reamy tomr@kapsgroup.com KAPS Group Knowledge Architecture Professional Services http://www.kapsgroup.com