How AI is impacting social data innovation
0 likes83 views
This document discusses how artificial intelligence is improving the analysis of social and customer data. It notes that language is complex with implicit meanings and that current text analytics have poor precision and recall. Machine learning, through supervised, unsupervised, and semi-supervised techniques, provides better analysis by discovering unseen patterns. Active machine learning replaces boolean searches with custom classifiers that achieve over 90% relevancy. Clear performance metrics make the models more transparent. Studies show these AI approaches achieve over 95% precision in social conversation analysis. They can help with brand equity, campaign measurement, customer experience, and linking social data to sales metrics.
More Related Content
Similar to How AI is impacting social data innovation (20)
Recently uploaded (20)
How AI is impacting social data innovation
- 1. AI Impact on Social and Voice of Customer Data Innovation Rob Key, CEO, Converseon
- 3. The same words mean different things An “unpredictable” movie is good, but “unpredictable” braking, not so much Much conversation is implicit ?? “I spent my entire lunch hour yesterday trying to exchange my American Airline ticket.” Sarcasm and slang abound “Lady Gaga killed it at halftime show.” Challenge: Poor Precision
- 4. ?? Is this tweet positive or negative? ?? It depends on whether you work for Verizon or Sprint ?? Often, your viewpoint is what makes something positive or negative, but most text analytics systems are not up to the challenge ?? “UberEats just launched in Ottawa but doesn't deliver to my house…I'm moving.” Every Signal, Target and Entity Must Be Captured and Analyzed Challenge: Poor Recall
- 5. A Case in Point.. ?? Estimated 60% precision in rules-based automated sentiment ?? 15-30% relevancy of dataset via boolean filtering ?? Low “recall” – document or sentence level
- 6. The Evolution for Computational Language Understanding Natural Language Processing 1950 1960 1970 1980 1990 2000 2010 * 2018 “Democratized AI” * Industry domain adaptation, ecosystem integration, model performance measurement and putting the power of machine learning directly into hands of citizen data scientists to build and improve their own models specific to organizational taxonomies and needs. Georgetown Experiment Micro World Ontologies Grammars Statistical Models Deep Learning
- 7. “Computers just don’t have the brains to solve some basic problems – unless we lend them ours.” Converseon lead scientist, Dr. Philip Resnik Machine Learning is Game-Changer in NLP
- 8. Semi-Supervised Machine Learning Keeps Human Intelligence in the Loop A combination of methods delivers the best results, including use of rules, client-specific domain expertise, and machine learning, which includes: Supervised Learning – depends on coded data i.e. what humans / analysts know Unsupervised Learning - uncovers unexpected patterns and language structure Semi-supervised Learning – utilizes a combination of approaches to capture human intelligence, while leveraging Machine Learning to: -? accelerate discovery of patterns -? building classifiers -? Confidence scores for each record enable efficient training
- 9. ("enterprise 2.0" OR "Social Biz" OR "social business" OR "#SocBiz" OR "#SocialBiz" OR "#socialbusiness" OR SocBiz OR SocialBiz OR socialbusiness OR "enterprise social business" OR "enterprise social network" OR "enterprise social grid" OR "social CRM" OR "social BPM") AND ("activity stream" OR "activity streams" OR "application development" OR "best practices" OR "business model" OR "business models" OR "content management" OR "crowd sourcing" OR "customer service" OR "disruptive technologies" OR "disruptive technology" OR "human resources" OR "information technology" OR "next generation social" OR "open social" OR "organizational culture" OR "predictive analytics" OR "product lifecycle management" OR "social analytics" OR "social mail" OR "social portal" OR "social portals" OR "social commerce" OR "customer experience management" OR "innovation management" OR "social app" OR "social application" OR "social applications" OR "social apps" OR "social business app" OR "social business application" OR "social business applications" OR "social business apps" OR "social collaboration" OR "social computing" OR "social learning" OR "social media" OR "social network" OR "social networks" OR "social platform" OR "social platforms" OR "social software" OR "social softwares" OR "social technologies" OR "social technology" OR "soft ware" OR "tool kit" OR "unified telephony" OR "analytics" OR "app" OR "application" OR "applications" OR "apps" OR "benefit" OR "benefits" OR "blog" OR "blogs" OR "bookmark" OR "bookmarks" OR "BPM" OR "collaboration" OR "collaborative" OR "commerce" OR "communication" OR "communities" OR "community" OR "compliance" OR "innovation" OR "connection" OR "connections" OR "CRM" OR "crowdsourcing" OR "engagement" OR "ERM" OR "information" OR "intelligence" OR "intelligent" OR "interaction" OR "interactive" OR "internet" OR "interoperability" OR "learning" OR "mail" OR "marketing" OR "mobile" OR "network" OR "nimble" OR "optimization" OR "optimized" OR "organization" OR "organizational" OR "organizations" OR "platform" OR "platforms" OR "portal" OR "portals" OR "productive" OR "productivity" OR "risk" OR "risks" OR "sales" OR "software" OR "softwares" OR "technologies" OR "technology" OR "tool" OR "toolkit" OR "tools" OR "transparency" OR "transparent" OR "value" OR "voIP" OR "wiki" OR "collective intelligence" OR "customer self service" OR "social grid" OR "content" OR "social content" OR "insights" OR "social selling" OR "social communications" OR "collaborative communications" OR "human capital management" OR "business process management" OR "enterprise resource management" OR "enterprise resource planning") 9 Active Machine Learning Replaces Booleans 15% relevancy 92% relevancy
- 10. 10
- 11. 11 Machine Learning “Custom Classifiers” Enable Broad and Deep Adoption Advocacy Customer Care Emotion Brand Attributes Unmet Needs Intensity Customer Journey Sentiment Purchase Intent Discovers patterns not visible To human analysis
- 12. 12 Essential: Clear Model Performance Scoring to Demystify the “Black Box” Precision
- 13. Peer-Reviewed Studies Demonstrate Predictive Capabilities of AI- Powered Social Data 13 Professor Wendy Moe and David Schweidel, conducted analysis of social conversation versus offline brand tracking using Converseon data. . New WOMM Media Mixed Modeling Study (utilizing Converseon data) 95% precision + 85% Relevancy + High Recall
- 14. Opening up the Value of this Data to New Critical Areas Brand Equity Campaign Measurement Brand Crisis & Resilience Product Launch Advocacy Media Impact on your Brand Social KPI impact on sales: sentiment, emotion, advocacy Mix Model / Predictive Analytics Customer Experience Track Facets that move needle
- 15. If interested in a free white paper on how AI is being applied to social and related Voice of Customer text data, simply email me at rkey@converseon.com 15 Thank You!