KMWorld 2024 - Butterfly Effect: Taxonomy and Ontology as AI Catalysts in Enterprise Learning
0 likes353 views
The document discusses the use of taxonomy and ontology in artificial intelligence (AI) to enhance enterprise learning content management. It provides case studies demonstrating successful implementations at organizations, highlighting the creation of intuitive taxonomies and ontologies that improve content findability, reduce redundancies, and facilitate personalized learning experiences. Best practices for developing these systems are also outlined, emphasizing the importance of user needs, scalability, and effective change management.
KMWorld 2024 - Butterfly Effect: Taxonomy and Ontology as AI Catalysts in Enterprise Learning
- 1. BUTTERFLY EFFECT: TAXONOMY AND ONTOLOGY AS AI CATALYSTS Tatiana Baquero Cakici & Sara Mae O’Brien-Scott KM World Taxonomy Boot Camp - November 2024
- 3. Agenda Federal AI Applications for Learning Content 1 2 Case Studies 3 Practical Considerations 4 Change Enablement and Adoption
- 4. Tatiana Baquero Cakici (she/her) Senior KM Consultant and Taxonomist Sara Mae O’Brien-Scott (she/her) Senior Semantic Engineering Consultant Presenters
- 5. EK At A Glance ESTABLISHED 2013 – PARTNERING WITH OUR CLIENTS TO TRANSFORM KNOWLEDGE INTO ACTION. SCALING TO OUR CLIENT’S NEEDS How We Can Engage Supporting at Any Stage DELIVERING VALUE ACROSS THE GLOBE HEADQUARTERED IN WASHINGTON, DC, USA PRESENCE IN BRUSSELS, BELGIUM DRIVING INDUSTRY THOUGHT LEADERSHIP Largest Open Knowledge Base Top-Ranked KM Podcast - Knowledge Cast KMWorld Reality Award Winner 2024 100 Companies That Matter in KM 2015-2024 AI 100 Trailblazer 2020-2024 Top-Selling Book on KM - Making Knowledge Management Clickable CONNECTING ALL ORGANIZATIONAL KNOWLEDGE ASSETS PRODUCTS KNOWLEDGE CONTENT INFORMATION DATA PEOPLE COLLABORATING TO ACHIEVE OUR CLIENT’S GOALS CONSISTENTLY AWARDED INC. BEST PLACES TO WORK BASED ON EMPLOYEE SATISFACTION SURVEYS
- 7. Common Challenges for Learning Content ● User experience ○ Findability: Difficulties connecting with content when users need it most ○ Discoverability: Lack of clear pathways through relevant content ● Content Management ○ Content Coverage: Identifying gaps in content ○ Duplicate Content: Identifying redundancies and duplications across content
- 8. ENTERPRISE KNOWLEDGE AI Applications for Enterprise Learning
- 9. AI Readiness in Learning Content To prepare content for AI: Apply taxonomy and ontology to specify relationship and structure, and aid in the clean up and the governance of the content. Model the knowledge domain Audit, clean up and de-duplicate content Add structure and standardization Componentize the content Build lifecycle management into governance policies and workflows
- 10. CASE STUDIES For Transforming Learning Content Experience with AI
- 11. Global Retail Organization Business Case The organization’s learning team struggled to search for, find, and deliver content stored in multiple locations to store employees. The were unable to identify obsolete content, outdated content, or gaps in their learning content, making content management inefficient. Beyond impacts to the learning team, store employees faced challenges to learning new roles and skills as it was difficult to find relevant learning content across the intranet and multiple learning apps. The Solution CREATED: Intuitive taxonomy and ontology for content. IMPLEMENTED: Learning Content Database that leverages the taxonomy and ontology for tagging and search. TRAINED: Project team empowered to manage and evolve the taxonomy. The Results The scalable solution resulted in Time saved searching for content 50% Reduction in time to tag content with metadata Increased ability to consistently track gaps and redundancies in content EK partnered with learning experts, content creators, and other members of the learning team to design a standardized taxonomy and ontology for tagging and relating learning content. The resulting models focused on store employee language and user needs to ensure that learning could be found more efficiently, increasing daily productivity and satisfaction. Up to @EKCONSULTING
- 12. . The Learning Content Database EK built the Learning Content Database to enable an enterprise view of learning content. Regardless of the system in which the learning content was created, it can be found from a single interface. Making this change to the learning content enables: ● personalized learning paths for learners ● enhanced content analysis and management @EKCONSULTING
- 13. Leadership at a global bank recognized that their employees were unable to leverage the organization's knowledge capital because it wasn't intuitively findable. The bank hoped to automatically assemble and proactively deliver targeted information to appropriate individuals. It was important to bank leadership that information was delivered at the moment of decision making and that time spent searching was minimized. The solution was built leveraging an ontology and knowledge graph, that collects organizational content, user context, and project activities. EK developed a unified, multilingual enterprise taxonomy for the global bank which is automatically and consistently applied to all relevant content to power highly relevant content recommendations. DESIGNED: Semantic model with organizational context and knowledge. INTEGRATED: Project content with context to drive knowledge sharing. DELIVERED: Custom content, context, and answers to users at their time of need. EK developed a multifaceted solution that achieved: Established a multilingual taxonomy with 8 systems Leadership calling the initiative “the most forward-thinking project in recent history” Delivered content with context to people at their time of need across EK developed a Semantic Hub (Findability) to ingest content from eight data sources and use machine learning and artificial intelligence to auto-tag content consistently and efficiently. Content is then assembled and intelligently delivered to five different front-end user experiences. 4 languages 4k+ terms for @EKCONSULTING Global Bank Business Case The Solution The Results
- 14. High-level Ontology Model Topic industry region country topic located in related to Annotation frequency score has annotation concept concept concept Location Country date publication date Content concept Content Type content type Industry
- 15. . The Semantic Hub In adding taxonomy and ontology to the organization’s content, the Bank was ready for a system that delivers highly contextualized, relevant content through multiple delivery channels including: ● Relevant content recommendations when scheduling a meeting ● Recommendations widgets which can be embedded on any page throughout the organization’s Intranet ● A corporate chatbot ● User-subscribed alert emails ● A search application, pooling content from multiple sources @EKCONSULTING
- 17. Best Practices to Define Taxonomies and Ontologies for AI Projects Identify Primary Personas and Use Cases Persona Type Role Taxonomy/ontology goals Taxonomy/ontology interactions Gain a greater understanding of the business challenges, goals, and motivations of end users to ensure the taxonomy/ontology focuses on improving their core daily activities. Reuse Standards and Existing Vocabularies Ensure consistency, interoperability, and efficiency of taxonomies and ontologies. Develop structures that facilitate data integration across systems and improve AI accuracy in understanding and processing information. Dublin Core SKOS RDF RDFS XML URI IRS OWL …
- 18. Considerations for Taxonomies and Ontologies used in AI Use Cases Scalability Flexibility to account for emerging needs and adapt without undue complexity. Granularity The right level of detail, strike a balance between efficient high level models and overly detailed and inefficient Interoperability The use of standards supports integration across source systems and formats. Alternative Labels Inclusion of synonyms and alternative terms to consider multiple end-user perspectives Multilingual Support Support for multiple taxonomies to provide recommendations and chatbot responses to global audiences. Behavioral Signals Inclusion of behavioral and user intent terms to help AI solutions adjust recommendations and responses to take user engagement patterns into account Machine Learning Enablement Alignment with AI/ML training models to support accurate decision making and flexibility to evolve based on ML driven insights. Audience and Personalization Categories and relationships that support personalization of content recommendations and chatbot responses.
- 20. Change Management: Guiding the Ripple Effect Clear Communication Strategy Establish the “What’s in it for me?” for leadership, key stakeholders, and priority end users. Align change plans and communications to business goals and strategic objectives. Document processes, intentions, and decisions … Start Small and Iterate Start with pilot projects that utilize a focused implementation in a controlled environment. Incorporate feedback from pilot project participants and gradually expand the scope. Training Train key stakeholders and priority end-users on the following as needed: Semantic Layers New Processes New AI Technologies
- 21. Thank you! Sara Mae O’Brien-Scott (she/her) Senior Semantic Engineering Consultant Enterprise Knowledge (EK) https://www.linkedin.com/in/saramaeobrienscott/ sobrien-scott@enterprise-knowledge.com Tatiana Baquero Cakici (she/her) Senior KM Consultant and Taxonomist Enterprise Knowledge (EK) https://www.linkedin.com/in/tatiana-baquero-cakici-pmp-610123/ tcakici@enterprise-knowledge.com