Overview of Taxonomies and Artificial Intelligence
0 likes•945 views
The document provides an overview of the role of taxonomies in artificial intelligence (AI), emphasizing their definition, creation, and implementation for tagging, classification, and enhancing findability. It discusses the integration of AI technologies like machine learning and natural language processing in developing and analyzing taxonomies, alongside the use of generative AI for term categorization and concept extraction. Finally, it highlights how taxonomies support various AI applications, including improved search mechanisms and recommendation systems.
Overview of Taxonomies and Artificial Intelligence
- 1. An Overview of Taxonomies and AI Heather Hedden Senior Consultant, Taxonomy and Ontology Enterprise Knowledge, LLC CILIP K&IM Group Artificial Intelligence Webinar Series: The promise and the perils 30 January 2024
- 2. ENTERPRISE KNOWLEDGE Outline Introduction to Taxonomies AI for Tagging and Classification with Taxonomies AI for Taxonomy Development AI for Taxonomy Analysis and Improvement Taxonomies for Supporting AI Applications
- 3. Introduction: What is a Taxonomy? A knowledge organization system that is… 1. Controlled: A kind of controlled vocabulary, based on unambiguous concepts, not just words (things, not strings). 2. Organized: Concepts are organized in a structure of hierarchies, categories, or facets to make them easier to find and understand. Controlled Organized
- 4. Introduction: What is a Taxonomy For? ⬢ Concepts are used to tag/categorize content to make finding and retrieving specific content easier. ⬢ Supporting better findability than search alone. ⬢ The taxonomy is an intermediary that links users to the desired content. Taxonomy implementation: focus on controlled Taxonomy implementation: focus on organized
- 5. Introduction: How are Taxonomies Created? ⬢ The taxonomy needs to be designed to suit both the users and the content. Content Taxonomy Users From Bottom Up: ⬢ Analyze the content ⬢ Conduct content audit ⬢ Extract terms from content From Top Down: ⬢ Interview stakeholders ⬢ Brainstorm in workshops and focus groups ⬢ Get users’ terms from search logs
- 6. AI for Tagging and Classification ⬢ Manual tagging and classification has limitations. Not scalable. ⬢ AI methods, existing since the 1990s, have become more common. ⬢ Not all “auto-tagging” uses AI. Rules can also be written to auto-tag. ⬢ Human review remains a feature. AI-based auto-tagging and auto-classification makes matches based on a linguistic, logical, or mathematical profile it expecets and recognizes. Technologies Include: ⬢ Machine learning (ML) and deep neural networks: Use mathematical analysis to find patterns that match known properties (text or images). ⬢ Named entity recognition: Matches proper nouns mentioned in text. ⬢ Semantic analysis: Locates concepts referenced within the content. ⬢ Natural language processing (NLP): Analyzes sentences.
- 7. AI for Taxonomy Development ⬢ Technologies for auto-tagging can also be used for extracting terms as candidate concepts for a taxonomy. ⬢ Machine learning and NLP enable term extraction from a “corpus” of documents. ⬢ Extracted terms are manually reviewed to be added to the taxonomy. ⬢ Software exists as stand-alone tools or features of taxonomy management systems. Creating hierarchical “is a” relationships can also be automated. However… ⬢ Lacking any input from users (interviews, surveys, etc.), results are not user friendly, but may be used in non-displayed taxonomies, such as for search support/SEO.
- 8. Generative AI for Taxonomy Development Using technologies such as ChatGPT with large language models (LLMs): ⬢ Request to put a list of terms (such as from term extraction) into categories. ⬢ Request suggested narrower concepts. ⬢ Request alternative labels (synonyms), including for certain audiences. Do not ask ChatGPT to “create” a taxonomy. It might violate copyrights.
- 9. AI for Taxonomy Analysis and Improvement ML, NLP, and Entity Recognition: Corpus Analysis ⬢ Rather than extract new, candidate terms, identify taxonomy concepts mentioned in the content. ⬢ Similar technology to auto-tagging, but may use other, similar content. ⬢ Identify high and low frequency occurring concepts.
- 10. AI for Taxonomy Analysis and Improvement Generative AI: Generating SPARQL Queries ⬢ For various analysis of a taxonomy built on SKOS standard. ⬢ For customized reports, that the taxonomy management software does not provide as a preset option.
- 11. Taxonomies Supporting AI Applications Taxonomies support applications, including: ⬢ Enhanced search and insights ⬢ Recommendation systems ⬢ Natural language question-answering ⬢ Chatbot support ⬢ Generative AI & LLM improved results (with Retrieval Augmented Generation/RAG) By providing: ⬢ Synonyms, homonyms, antonyms, etc. - for normalization and disambiguation ⬢ Hierarchies - for context ⬢ Components of an ontology - for domain knowledge Which enables: ⬢ Matches based on relationships, profile and behavior ⬢ Query disambiguation ⬢ Query expansion / refinement
- 12. Resources ⬢ “The Role of Ontologies with LLMs,” (January 9, 2024) by James Midkiff, Enterprise Knowledge Blog. ⬢ “How a Knowledge Graph Supports AI: Technical Considerations,” (September 26, 2023), by Urmi Majumder, Enterprise Knowledge Blog. ⬢ “Expert Analysis: When should my organization use auto-tagging?,” Part 1 (July 19, 2022) and Part 2 (December 20, 2022) by James Midkiff and Sara Duane, Enterprise Knowledge Blog. ⬢ “ChatGPT and Generative AI for Taxonomy Development” (PPT), presented by Xia Lin, Taxonomy Boot Camp Conference, November 7, 2023. ⬢ “ChatGPT, Taxonomist: Opportunities & Challenges in AI-Assisted Taxonomy Development” (PPT), presented by Margie Hlava and Heather Kotula, Taxonomy Boot Camp Conference, November 7, 2023. ⬢ “Taxonomies and ChatGPT,” (May 29, 2023), by Heather Hedden, The Accidental Taxonomist Blog.
- 13. Q&A Thank you for listening. Questions? Heather Hedden Senior Consultant Enterprise Knowledge, LLC www.enterprise-knowledge.com [email protected] www.linkedin.com/in/hedden