AI-SDV 2021 - Marjorie Hlava - Semantic Search and Content Management – Case Studies in Successful Software Implementations
0 likes288 views
The document discusses the challenges faced by various organizations in content management and discovery, highlighting the need for improvements in search results and content quality. It presents AI-based solutions and automated tagging systems implemented by institutions such as the Metropolitan Opera and the American Society for Clinical Oncology to enhance content usability and accuracy. The case studies illustrate how these organizations achieved significant improvements in their content search and management processes.
AI-SDV 2021 - Marjorie Hlava - Semantic Search and Content Management – Case Studies in Successful Software Implementations
- 1. Semantic Search and Content Management Case Studies in Successful Software Implementations Marjorie M.K. Hlava Founder/President mhlava@accessinn.com
- 2. ü Content management and content discovery needed major improvements. ü User were not getting the results they needed. ü The content production team including editorial, managing editors – the whole team – could no longer cope with the volume and variety. ü Content quality was suffering. ü Brief discussions of each organization’s challenges set the stage for AI- based, human curated solutions. ü What worked, what didn’t, and the how and the why will be presented. What do PLI, Met Opera, ASCO, McGraw-Hill and PLOS have in common?
- 3. Who are they? üPracticing Law Institute onon-profit continuing legal education organization ochartered by the Regents of the University of the State of New York. oFounded in 1933 othe company organizes and provides CLE programs around the world. üHow to make their content discoverable üHow to reuse and cross reference training materials üTime crunches and staff overloaded already üLot of content so need automation to make this happen
- 4. First Auto-tag the Content When Ingested
- 5. Keep Track of Every Record and the Level of Tagging
- 6. Everyone Gets a Login to the Metadata Platform – Levels of Permission Differ
- 7. üTailored the taxonomy to the content üAuto-tagging on backfile and forward flow production üEnsure accuracy in tagging with QA program of spot checking and automatic monitors üEnriched data quality and coverage üAllowed creation of new CLE offerings through content reuse üSignificantly improved search for both staff and customers
- 8. Financed by programs and donations Database, DOS based implemented in 1980’s Performers use the archives database as their resume referral system Keep track of every performer in every scene of every opera ever preformed at The Met Time critical data input needed Archives mostly staffed by volunteers – make database easy to use Metropolitan Opera
- 10. Metropolitan Opera üAdd images and videos üConnect to costume database üSupport the archives website üAllow immediate entry of data after each performance in season üSubmit comments and corrections online üMake search fast, accurate, and easy
- 11. American Society for Clinical Oncology üWeb search inconsistent results üTaxonomy to support synonyms and variant term usage üInsure comprehensive search üType ahead in search üSort conference papers into tracks üTag to user profiles for better matching of talks to attendees üCreate meta-titles for better search SEO üUse in journal productions tracking üHelp match peer reviewers to potential papers
- 12. American Society for Clinical Oncology Metastatic 170,000,000 Stage 4 296,000,000 Invasive 92,000,000
- 13. Consistent Search Based on Taxonomic Metadata üSearch for oInvasive breast cancer oMetastatic breast cancer oStage IV breast cancer üGet a single search set result whatever term set you enter Funding – 2 Conferences -3862 Journals 2993
- 14. Well formed vocabulary control Rich Synonymy and cross references Deep automatic tagging Taxonomy – Tagging – Data Enriched
- 15. - Content Analysis - Term Identification - Multiple Taxonomies with Weighted Terms - Enriched Content (XML, Video Transcript, Excel) Automated Tagging - Taxonomy Development - AI Training Set - Classifying - Metadata Enrichment
- 16. Lauren Sapira Director, Sci/Tech Digital Products “ “ Before we did this upgrade project, we heard all kinds of complaints about our search. Since we relaunched, there have been no complaints at all!
- 17. Public Library of Science üTag every document on ingestion üWebsite search platform üThe search hierarchy is visible to 3 levels deep for enhanced search üUsers can vote on keywords used to train the system üIndex to the most specific level üEditors can view the entire hierarchy üSearch is keyword driven based on the taxonomy metadata üTrack every usage of data by searchers
- 18. Public Library of Science üNeed to keep track of all incoming papers as well as usage by customers üWhat new journals should we create, and which are not doing well? üWho uses what on the web site üWhat works, what does not üIf we change something, what is the impact?
- 19. Show keywords and allow users to suggest additions or deletions
- 20. Automatic term suggestions for incoming documents
- 21. Automatic weighting of taxonomy terms based on source content
- 22. Summary Content enrichment gives better data Use subject metadata for better search Don't launch at less than 85% accuracy -- Precision and recall combined Unstructured data is still structured and can be tagged Keep track of the data usage
- 23. About ALBUQUERQUE PHILADELPHIA Cambridge • Established in 1978 • Experts in semantic solutions • Over 300 clients on 4 continents • Over 2,000 project engagements • 95% YOY client retention • Data Harmony Suite • Award Winning • Patented Technology • Semantic Services • Managed Services • Project Services
- 24. Clients Publishing & Media Education Government Non-profits & Societies Health/Pharma Manufacturing & Retail
- 25. Services Overview We are experts in semantic solutions ü Metadata Creation and Enhancement ü AI Training Set Development ü Custom Data Classification ü Develop Controlled Vocabularies ü Abstract and Index Documents ü Automated Indexing ü Capture and Convert Data Not only do we build solutions that are compliant, in many cases, we built or influence the data standard!
- 26. Semantic Environment ü Concept Extraction ü Term Recommendations ü Sentiment Analysis ü Visualization Tools ü NLP Precision Monitoring ANALYZE ü Taxonomy Building ü Thesaurus Building ü AI Training-Set Building MODEL ü Abstracting ü Classifying ü Metadata Enrichment ü Inline Tagging ü Meta-Titles ENRICH
- 27. Marjorie MK Hlava Founder/President mhlava@accessinn.com (m): 505-975-5578 Change Search to Found We are the intelligence and the technology behind world-class semantic solutions