Let Search Power Your Intranet!
2 likes431 views
Ravi Mynampaty gives a presentation on implementing an enterprise search solution for an intranet. He discusses assembling content from various systems into a single search index after cleansing, supplementing, and harmonizing the data. This One True Collection approach provides federated search across systems and standardized metadata and fields. The talk covers architectural considerations for search, security, and demonstrates the solution.
Let Search Power Your Intranet!
- 1. Copyright © President & Fellows of Harvard College Let Search Power Your Intranet! Ravi Mynampaty
- 2. About Ravi A hustler making a living by pretending to know more about Enterprise Search than he actually does...
- 4. “I can live on a good compliment two weeks with nothing else to eat...” @RaviMynampaty
- 5. Why the heck should I listen to Ravi?
- 9. What's this talk about?
- 11. Data lakes
- 12. Warehouses
- 13. Silos
- 15. Cleanse
- 16. Assemble
- 17. Supplement
- 18. Harmonize
- 19. Content Search Index CMS Records Web pages RDBMS Assemble HarmonizeCleanse Supplement Learning Management Systems Service Management Systems CRM Systems etc.
- 21. Why?
- 25. Options What? VF ⇒ PF Replace variant forms (VF) with Preferred form (PF) E.g., All variant forms of “Apple...” ⇒ “Apple Inc.” Where? Source UISearch Index
- 29. Metadata PDF
- 33. Why?
- 36. Document Analytics (Popularity) Search Index Single Record - Postal code ⇒ City, State - Implicit metadata - Link depth
- 37. Supplemented Records Id <Record 1> URL www.hbs.edu/faculty/ Popularity 950218 LinkDepth 1 Id <Record 2> URL www.hbs.edu/mba/academic-experience/blog/post/hbs-global-initiative-research-centers Popularity 2493 LinkDepth 6
- 38. For the sake of Relevance! http://www.hbs.edu/search.aspx?q=finance&.. ...&bboost=sum( product (sub(10,LinkDepth),0.1), max(log(Popularity),1) ) &...
- 40. Standardized Field Names Record 3 Field: Webaddress Record 1 Field: URL Record 2 Field: Link Record 3 Field: HBSLink Record 1 Field: HBSLink Record 2 Field: HBSLink Search Index
- 42. Content Search Index CMS Records Web pages RDBMS Assemble HarmonizeCleanse Supplement Learning Management Systems Service Management Systems CRM Systems etc. OTC: One True Collection
- 43. Why?
- 44. Users
- 45. Federated Search One search box to rule them all Multiple search tool federation Harmony at the SERP level What we thought was the Holy Grail One True Collection (OTC) One search box to rule them all Single search index Harmony at the result level What the users wanted
- 48. Content Search Index Your Bank! ● Fast lookup ● Web services for all our content CMS Records Web pages RDBMS Assemble HarmonizeCleanse Supplement Learning Management Systems Service Management Systems CRM Systems ● Normalize data ● Remove special chars ● … ● Standard fields ● One True Collection ● ... ● Analytics (popularity) ● Postal Code⇒ city,state ● Implicit metadata ● LinkDepth ● ... Intranet Websites etc. Start making withdrawals !! Joins ● PDF Full-text ● Person record ● ... etc. CASH
- 50. WebServicesAPIs Solr CMS Web Legacy 1 ITAssets Legacy 2 Assets Collections Oracle DB Java DBImport Loader Web Services APIs Web Connector Solr XML XML Connector Informatica Web Connector Crawl Pages CMS Web Services APIs Web Loader Intranet, Apps, Portals Websites Websites HBS Search Service Enterprise Search Architecture
- 51. Some Architectural Considerations Search index design Hardware, Scalability Query optimization High Availability (HA), Disaster Recovery (DR) Analytics, Ongoing relevance tuning Security
- 53. Some Security Considerations Logged In/Out Repository level, Document level, Field level Group-based, Role-based, Individual-based Index-time vs. Query-time
- 56. Demo!