Taxonomy Development Knowledge Structures Tom Reamy Chief Knowledge Architect KAPS Group Knowledge Architecture Professional Services http://www.kapsgroup.com
Agenda Introduction Knowledge Structures Taxonomy Management Software Exercises Conclusion
Knowledge Structures List of Keywords (Folksonomies) Controlled Vocabularies, Glossaries Thesaurus Browse Taxonomies (Classification) Formal Taxonomies Faceted Classifications Semantic Networks / Ontologies Topic Maps Knowledge Maps
Knowledge Structures Lists of Keywords (Folksonomies) Folksonomy (also known as collaborative tagging, social classification, social indexing, and social tagging) is the practice and method of collaboratively creating and managing tags to annotate and categorize content. Folksonomy describes the bottom-up classification systems that emerge from social tagging.[1] In contrast to traditional subject indexing, metadata is generated not only by experts but also by creators and consumers of the content. Usually, freely chosen keywords are used instead of a controlled vocabulary.[2] Folksonomy (from folk + taxonomy) is a user generated taxonomy.
Knowledge Structures Controlled Vocabularies, Glossaries Lists with minimum structure Easy to develop Difficult to get value from Simple Reference resource Thesaurus Taxonomy-like Less formal BT, NT – also RT
Two Types of Taxonomies: Browse and Formal Browse Taxonomy – Yahoo
Two Types of Taxonomies: Formal
Facets and Dynamic Classification Facets are not categories Entities or concepts belong to a category Entities have facets Facets are metadata - properties or attributes Entities or concepts fit into one category All entities have all facets – defined by set of values Facets are orthogonal – mutually exclusive – dimensions An event is not a person is not a document is not a place. Facets – variety – of units, of structure Date or price – numerical range Location – big to small (partonomy) Winery – alphabetical Hierarchical - taxonomic
Knowledge Structures Semantic Networks / Ontologies Ontology more formal XML standards – OWL, DAML Semantic Web – machine understanding RDF – Noun – Verb – Object Vice President is Officer Build implications – from properties of Officer Semantic Network – less formal Represent large ontologies Synonyms and variety of relationships
Knowledge Structures: Ontology Instruments Music is a is a create Bluegrass Violins uses Musicians uses is a Violinists
Knowledge Structures Topic Maps ISO Standard See www.topicmaps.org Topic Maps represent subjects (topics) and associations and occurrences Similar to semantic networks Ontology defines the types of subjects and types of relationships Combination of semantic network and other formal structures (taxonomy or ontology)
Knowledge Structure: Topic Maps
Knowledge Structures Knowledge Maps No standards – applied at high level Ontologies plus / applied to specific environment Map of Groups – Content Stores – Purpose – Technology Add structure to each element Facet Structure – filter by group – content – purpose Strategic resource
Knowledge Structures: Which one to use? Level 1 – keywords, glossaries, acronym lists, search logs Resources, inputs into upper levels Level 2 – Thesaurus, Taxonomies Semantic Resource – foundation for applications, metadata Level 3 – Facets, Ontologies, semantic networks, topic maps Applications Level 4 – Knowledge maps Strategic Resource
Web 2.0 – No need for Taxonomies etc.? “Tags are great because you throw caution to the wind, forget about whittling down everything into a distinct set of categories and instead let folks loose categorizing their own stuff on their own terms." - Matt Haughey - MetaFilter Tyranny of the majority - worst type of central authority More Madness of Crowds than Wisdom of Crowds “Things fall apart; the center cannot hold; Mere anarchy is loosed upon the world,… The best lack all conviction, while the worst Are full of passionate conviction.” - The Second Coming – W.B. Yeats
Advantages of Folksonomies Simple (no complex structure to learn) No need to learn difficult formal classification system Lower cost of categorization Distributes cost of tagging over large population Open ended – can respond quickly to changes Relevance – User’s own terms Support serendipitous form of browsing Easy to tag any object – photo, document, bookmark Better than no tags at all Getting people excited about metadata!
Folksonomies – Problems and Limits Folksonomies don’t compare with taxonomies or ontologies Serendipity browsing is small part of search Limited areas of success – popular sites are popular Quality Content – finance, science, etc – not good candidates No mechanism for improving folksonomies Scale – Too Big (million hits) – Too Little (200 items) – Amazon and LibraryThing Need intrinsic value of tagging – not tagging for better tags Bad Tags - idiosyncratic or too broad, errors, limited reach Most people can’t tag very well – learned skill
Del.icio.us Tags Design blog software music tools reference art video programming webdesign web2.0 mac howto linux tutorial web free news photography shopping blogs css imported education travel javascript food games Development inspiration politics flash apple tips java google osx business windows iphone science productivity books toread helath funny internet wordpress ajax ruby research humor fun technology search opensource Photoshop media recipes cool work article marketing security mobile jobs rails lifehacks tutorials resources php social download diy ubuntu freeware portfolio photo movies writing graphics youtube audio online
Del.icio.us - Folksonomy Findability Too many hits (where have we heard that before?) Design – 1 Mil, software – 931,259, sex – 129,468 No plurals, stemming (singular preferred) Folksonomy – 14,073, folksonomies – 3,843, both – 1,891 Blog-1.7M, blogs – 516,340, Weblog- 155,917, weblogs – 36,434, blogging – 157,922, bloging – 697 Taxonomy – 9.683, taxonomies – 1,574 Personal tags – cool, fun, funny, etc Good for social research, not finding documents or sites How good for personal use? Funny is time dependent
Library Thing Book people aren’t much better at tagging High level concepts – psychology (55,000), religion (120,000), science (101,000) Issue – variety of terms – cognitive science – need at least 40 other tags to cover the actual field of cognitive science Strange tags – book (19,000) – it’s a book site? Combination of facets and topics Facets – Date (16th century, 1950’s, 2007) // Function (owned, not read) // Type (graphic novel, novel) // Genre (horror, mystery) Topics – majority like Del.icio.us
Library Thing – Book on Neuroscience 1) (Location: dining room)(1) biological(1) biology(8) box74(1) Brain(1) brain research(1) brains(1) cognitive neuroscience(1) cognitive science(1) consciousness(1) currently reading(1) HelixHealth(1) kognitionswissenschaft(1) medical(1) medicine(1) neuroscience(19) non-fiction(5) partread(1) Psychology(4) Science(10) textbook(10) theory(1) Too General: Science, Psychology, biology, textbook Too specific: Location: dining room, box74 Facets: currently reading, partread
Better Folksonomies: Will social networking make tags better? Not so far – example of Del.icio.us – same tags Quality and Popularity are very different things Most people don’t tag, don’t re-tag Study – folksonomies follow NISO guidelines – nouns, etc – but do they actually work – see analysis Most tags deal with computers and are created by people that love to do this stuff – not regular users and infrequent users – Beware true believers!
Browse Taxonomies: Strengths and Weaknesses Strengths: Browse is better than search Context and discovery Browse by task, type, etc. Weaknesses: Mix of organization Catalogs, alphabetical listings, inventories Subject matter, functional, publisher, document type Vocabulary and nomenclature Issues Problems with maintenance, new material Poor granularity and little relationship between parts. Web site unit of organization No foundation for standards
Formal Taxonomies: Strengths and Weaknesses Fixed Resource – little or no maintenance Communication Platform – share ideas, standards Infrastructure Resource Controlled vocabulary and keywords More depth, finer granularity Weaknesses: Difficult to develop and customize Don’t reflect users’ perspectives Users have to adapt to language
Faceted Navigation: Strengths and Weaknesses More intuitive – easy to guess what is behind each door 20 questions – we know and use Dynamic selection of categories Allow multiple perspectives Trick Users into “using” Advanced Search wine where color = red, price = x-y, etc.. Weaknesses: Difficulty of expressing complex relationships Simplicity of internal organization Loss of Browse Context Difficult to grasp scope and relationships Limited Domain Applicability – type and size Entities not concepts, documents, web sites
Dynamic Classification / Faceted navigation Search and browse better than either alone Categorized search – context Browse as an advanced search Dynamic search and browse is best Can’t predict all the ways people think Advanced cognitive differences Panda, Monkey, Banana Can’t predict all the questions and activities Intersections of what users are looking for and what documents are often about China and Biotech Economics and Regulatory
Varieties of Taxonomy/ Text Analytics Software Taxonomy Management Text Analytics Auto-Categorization, Entity Extraction Sentiment Analysis Software Platforms Content Management, Search Application Specific Business Intelligence
Vendors of Taxonomy/ Text Analytics Software Attensity Business Objects – Inxight Clarabridge ClearForest Data Harmony / Access Innovations Lexalytics Multi-Tes Nstein SchemaLogic Teragram Wikionomy Wordmap Lots More
Why Taxonomy Software? If you have to ask, you can’t afford it Spreadsheets Good for calculations, days of taxonomy development over (almost) Ease of use – more productive Increase speed of taxonomy development Better Quality – synonyms, related terms, etc. Distributed development – lower cost, user input (good and bad)
Text Analytics Software – Features Taxonomy Management Functions Entity Extraction Multiple types, custom classes Auto-categorization – Taxonomy Structure Training sets – Bayesian, Vector space Terms – literal strings, stemming, dictionary of related terms Rules – simple – position in text (Title, body, url) Boolean– Full search syntax – AND, OR, NOT Advanced – NEAR (#), PARAGRAPH, SENTENCE Advanced Features Facts / ontologies /Semantic Web – RDF + Sentiment Analysis
Conclusion Variety of information and knowledge structures Important to know what will solve what Taxonomies and Facets are foundation elements Build higher levels based on lower levels Glossaries to Taxonomies Taxonomy to Ontology / faceted navigation Important to have good taxonomy and text analytics software (spreadsheets are OK for first draft) Web 2.0/Folksonomies are not the answer
Resources Books Software Web Sites Women, Fire, and Dangerous Things George Lakoff Knowledge, Concepts, and Categories Koen Lamberts and David Shanks The Stuff of Thought – Steven Pinker Software Tools & Techniques (Taxonomy Boot Camp) Web Sites Taxonomy Community of Practice: http://finance.groups.yahoo.com/group/TaxoCoP/
Questions? Tom Reamy tomr@kapsgroup.com KAPS Group Knowledge Architecture Professional Services http://www.kapsgroup.com