1. Leveraging Taxonomy & Metadata For Superior Search Relevancy Achieve greater search relevancy with content structure Speaker: Marko Hurst
2. Me Consultant, Author, & Speaker User Experience / Experience Design Web Analytics Search Background: 14 years experience Search Systems, Data Analysis, Enterprise Applications, Websites, Mobile, Web. 2.0 Independent, agency, & consulting firms National & regional lead for UX, Strategy, & Web Analytics practices Industries: Government, Media, eCommerce, Financial Services, Automotive, Technology, Mobile, CPG Contact: MISI, Engagement Manager:mhurst@misicompany.com Read my Blog:MarkoHurst.com“Insightful Analytics” Follow me on Twitter:MarkoHurst
3. Me Book: Search Analytics - Conversations With Your Customers Anticipated release: late 2010 Book website:RosenfeldMedia.com/books/SearchAnalytics Co-Author: Lou Rosenfeld Speaker: Keynote North America & Europe eMetrics Marketing Optimization Summit Search Marketing Exchange, SMX Usability Professional Association, UPA Various Digital Asset Management Technology Marketing Corporate Government Agencies
5. Audience Survey Question 1 Who uses only an algorithm (does not leverage a formal taxonomy or metadata structure) for search results? It’s OK, I’m here to help
6. Audience Survey Question 2 Who uses a taxonomy to aid search relevancy? Sweet! – You’re ahead of most
7. Audience Survey Question 3 Who uses a ontology to aid search relevancy? I salute you! You will find Nirvana here.
8. What We’re Going To Cover Definition of Common Terms Taxonomy and Search Metadata and Search Ontology and Search Q&A
10. Definition: Taxonomy Snippet A parent / child hierarchal relationship between two or more items English Knowledge map that allows users to access relevant objects, ideas and/or experts quickly and efficiently Taxonomies classify domains of knowledge and show the hierarchical relationships between categories, sub-categories and values within categories Geek Speak Taxonomy is the practice and science of classification and comes from the Greek “taxis” – “order" (or arrangement or division) and "nomos", meaning law or science Taxonomies, which are composed of taxonomic units known as taxa (singular taxon), are frequently hierarchical in structure, commonly displaying parent-child relationships
11. Taxonomy: Sample Screenshot - DAM System http://www.day.com/content/day/en/products/digital_asset_management/features/_jcr_content/par/image.img.gif
13. Definition: Metadata Snippet “Data about data” of any sort in any media (paper-based or electronic media). English Metadata describes how and when and by whom a particular asset was collected, and how the asset is formatted asset in order to provide access to the asset Metadata is text, voice, or image that describes what the audience wants or needs to see or experience Geek Speak In its broadest sense, “metadata” can be used to describe information structures Metadata is a summary of the form and content of a resource, i.e. books: titles, authors, publishers, ISBN, etc. Usually includes information about the intellectual content of the image, digital representation data, and security or rights management information
14. Metadata: DAM System Assigning Metadata http://dev.day.com/content/docs/en/cq/current/dam/how_to_edit_metadata/_jcr_content/par/image_6.img.png/1258559070876.png
15. Definition: Ontology Snippet Form associative relationships between two or more items English Metadata describes how and when and by whom a particular asset was collected, and how the asset is formatted asset in order to provide access to the asset an explicit formal specification of how to represent the objects, concepts and other entities that are assumed to exist in some area of interest and the relationships that hold among them and describes rather than the hierarchy, the relationship between entities Geek Speak Ontologies resemble faceted taxonomies but use richer semantic relationships among terms and attributes, as well as strict rules about how to specify terms and relationships Because ontologies do more than just control a vocabulary, they are thought of as knowledge representation Ontologies can represent complex relationships between objects, and include the rules and axioms missing from semantic networks
20. Surfacing Content With Search Two dimensions for surfacing content within search: Semantic and Taxonomic Semantic / Text paradigm Pertains to search only Search (pull) relies upon textual matching and semantic algorithms to surface relevant content Search engine derives semantics from phrases and words in unstructured content and from field-definition in structured content Taxonomic paradigm Pertains to search (pull), personalization, and customization (push) Utilizes a taxonomy to surface relevant content. Search interrogates taxonomy and ontology (associative relationships) Be aware Both approaches have advantages and disadvantages Both approaches have significant challenges – there are NO easy options! Possible to incorporate both approaches in a single search solution in effect creating two concurrent searches
21. Semantic vs. Taxonomy Based Search Semantic/Free Text Search Searches against the content of a database repository (i.e. involves only two steps: search contents) Uses only the keyword(s) entered into the search engine Where these is an exact match, a result is returned This gives results that are less expansive, less controlled and often, less relevant...
22. Semantic vs. Taxonomy Based Search Taxonomy Based Search Searches against the metadata associated with the content stored in a database repository (i.e. involves three steps: search metadata contents) The metadata (NOT the search itself) is mapped against the contents of a database repository Can match the user’s search word entered into the search engine with synonyms mapped in the taxonomy (e.g. “"P.D. 533" maps to "Presidential Decree No. 533” and “The anti-Cattle Rustling Law of 1974”). Matches can be made between a user's query and related terms mapped in the taxonomy, e.g. “P.D. 533” might be mapped to "Cattle Theft"; "Felony" to "Legislation”)
23. Semantic vs. Taxonomy Based Search Taxonomy Based Search (con’t) Search results can be prioritized and categorized by filtering for pages and/or documents associated with specific search terms e.g. "best bets” When there is ambiguity, can ask users to refine their searches by providing “Did you mean...?” feedback This gives results that are more expansive and relevant…
24. How Taxonomy Aids Searching A taxonomy aids searching by… Restricting searches within a finite category or set of categories e.g. a search for “Farm Bill" will be restricted to the category "Legislation” Expanding searches to higher (parent), lower (e.g. child) or equivalent (e.g. sibling) categories e.g. a search for “"Cattle Poaching" would search across "Legislation", "PB 553", “SB 1163” "Felony" “Products”, etc.
25. Restricting / Expanding Search With Taxonomy Parent node Sibling node Search for “Farm Bill” Child node Child node
26. How Taxonomy Aids Searching (con’t) Provide “Did you mean?” feedback to users to refine searches e.g. a search for “Livestock” might return: “Did you mean Livestock Health, Livestock Management, Livestock Legislation or Supplies & Equipment?” Search against synonyms (i.e. alternate terms) e.g. a query against the acronym “DDT" would map to “dichlorodiphenyltrichloroethane”, and search in the "Pesticides" category Search against related terms e.g. “Pesticide" is a term that exists in both "Crop Plants" and "Products” Because this relationship is known (and mapped) in the taxonomy, searches on one usage will also return "hits" on the other Search on obscure or obtuse relationships e.g. Paul Hermann Müller, Rachel Carson, environmental movement, and the book Silent Spring can all be mapped
27. How Taxonomy Aids Searching (con’t) Allows for easier auto-complete / type ahead functionality Serves as a short‐cut Helps users to avoid unnecessary typing Assists with spelling May suggest related or more specific queries (that begin with or include that word or phrase
29. Summary Two dimensions for surfacing content within search: Semantic and Taxonomic Semantic is pull only Taxonomic is push & pull Both have pros & cons The best search results can typically be achieved by using both Taxonomy is a parent / child relationship between two or more items Taxonomic search Allows for use of synonyms & mapping to related & obscure relationships Allows for expanding and restricting of content by moving up (parent), moving across (sibling), or moving down (child) nodes within the taxonomy Many benefits / features can be used within the interface using taxonomy Did You Mean?, Auto-Suggest, Best Bets, etc. Metadata Metadata is “data about data” of any sort in any media Taxonomies provide an inherent level of metadata that is not possible otherwise Leveraging metadata frameworks (Dublin Core, PRISM, etc) allow for standard methods of Ontology Associative relationship between two or more items Synonyms, controlled vocabularies, metadata, etc can be mapped taxonomy items for greater expansion & contraction of related content