This is a whirlwind introduction to taxonomy basic principles: the how, the why and what you need to get the job done quickly and correctly. Hear about the standards and references available, as well as the basic building blocks needed to create a well-formed and versatile taxonomy. This session is a good foundation for those new to taxonomy design to help understand the rest of the program. Grab your breakfast and join our expert as she gets you ready for an intensive Taxonomy Boot Camp!

Conference program chair, Stephanie Lemieux, welcomes attendees to the start of Taxonomy Boot Camp 2017!

Ever wonder how—and why—Netflix gets so super- specific with its categories? It may seem like the work of an over-eager editorial staff. In actuality, it’s a highly deliberate—and heavily tested—tag-based taxonomy, designed to suit 100 million users and scale for any movie or TV show in the world. Hastings delves into one of the company’s more successful experiments in merchandising, and reveals the editorial strategies involved in putting a human face on its seemingly infinite, algorithmic personalization.

This interactive session covers the essential role that taxonomies play in supporting search, navigation, content management, and knowledge management processes for both internally and externally focused enterprise taxonomy projects. Participants gain an understanding of taxonomies, their roles in supporting organizational business goals, and the process for designing and building them to meet a wide range of needs. In particular, participants get an understanding of the breadth of inputs needed to design durable and sustainable taxonomies.

What can you do when a hierarchy of terms just isn’t enough? Etsy is tackling this issue by introducing structured product data to our taxonomies. Beginning with a brief history and overview of taxonomy at Etsy, hear about the metadata fields created for product categories: how they were defined, what they can be used for, and how they were modeled in their own internal taxonomy platform. Explore the benefits and challenges of integrating structured data into an existing taxonomy, especially one as wild and unique as Etsy’s.

Sandia National Labs is a multi-program laboratory with more than 10,000 employees working on a wide range of projects in various subject areas. Its Analytics for Sandia Knowledge (ASK) Expertise Finder application can be used for strategic staffing, visualization of expertise trends, and for identifying networks of collaborators. Using machine learning and NLP algorithms on information that is produced through normal work processes, the application is self-maintaining. Hear about an extension of this application that generates and organizes expertise for individual Sandia organizations, and how the data is organized and displayed using meaningful subject areas—the Sandia National Laboratories’ Subject Category Guide (SCG).

Last year, Consumer Reports launched its first enterprise-wide taxonomy and selected a taxonomy management tool. A year later, Fleshler shares the successes, challenges, and lessons learned as Consumer Reports moved into the initial implementation phase. Hear how Consumer Reports is implementing the CR Taxonomy as the backbone of internal performance analytics, specifically web metrics, content tagging for online and print articles, and employee time allocation. Also learn how it is leveraging results from these and other sources as part of a movement toward a taxonomy- driven analytical engine that drives future functionalities on its website, including search, personalization, and navigation.

To combat variations in professional jargon and bring uniformity to the practice of radiology and imaging, the RSNA has created RadLex, an “official” controlled vocabulary for the profession. It has been adapted as the basis for structured radiology reporting, a national radiation dosage registry, common data elements, and a manual of uniform imaging protocols and medical billing codes. Versions of RadLex have also been adapted for usage in semantic enrichment for online publishing. Hear about how the RadLexbased taxonomy is being converted for use in discovery tools for RSNA’s developing digital repository, which will hold studies, journal articles, images, DICOM stacks, and radiological cines.

Enterprise search is hard. The information employees need to be productive is often spread across multiple locations on the web, professional research data services, and work data each with their own way to discover what you need. Searching across this family of sourc-es takes a lot of time, and it’s often hard to get a single global view into a topic across multiple sources. Hear how customers in academia and industry have added a new search experience within their organizations.

To ensure maximum success and value of the taxonomy for the organization, it is essential to hear from as many potential users and stakeholders as possible during the taxonomy design process. Capturing quantitative and qualitative feedback from end users is critical to gain consensus and make adjustments to the taxonomy during development. Learn about taxonomy and usability validation techniques that have effectively corroborated the usefulness and intuitiveness of a taxonomy design, the strengths and weaknesses of each technique, a set of tools that supports the process, and real cases studies from private and public sectors.

Comcast Internal Communications needs an enterprise taxonomy of business terminology to tag enterprise portal content and surface it efficiently in its new enterprise search engine. Hear how Carilla is using his experience creating full-service enterprise taxonomy capability at a multinational pharmaceutical to develop a new road map to guide Comcast in implementing this baseline taxonomy, along with all of the information architecture surrounding it (e.g., principles, standards, guidelines, governance) to create value and be able to grow as enterprise demands increase.

You have done the research, painstakingly developed a foundational taxonomy, presented it to the team and are convinced your content categorization woes are over and the world is a better place already. Three months in though, enthusiasm has waned and adoption is low among the team. Why is this so, and what can you do about it? Roux guides you through some common issues with taxonomy adoption and provides practical ways to overcome them.

A global Fortune 500 company needed an experience marketing platform that would support any number of business units marketing any number of products to any number of customers across multiple channels with an unknown mix of static and dynamic content and complex personalization yet to be determined—because the company knew it was in transition, the platform would need to evolve without any new development. How do you design a sustainable information architecture when organization, labels, navigation, and metadata are guaranteed to change? Hear lessons from designing this and other flexible organizational systems, and learn approaches to use when architecting sustainable, complex, enterprise platforms.

The requirements that get published in RFPs reveal how an organization sees its needs and its knowledge of how to conduct a taxonomy project with the best chance of success. However, badly scoped projects can reveal ignorance, common misconceptions, and the failure to manage stakeholders well. Lambe takes an entertaining look at some terrible examples of published requirements and the underlying errors and misconceptions about taxonomy projects that they reveal. Anybody involved in taxonomy projects can learn about how to scope a project for success from these examples.

Machine learning algorithms can complement human intelligence with their ability to extract patterns from vast amounts of information rapidly. Algorithms that learn from reference text corpora can provide taxonomists with valuable insights: How complete is our taxonomy? Which areas need to be extended? Which are overrepresented? Hear how taxonomists can interact with a recommender system based on corpus learning. Blumauer discusses where the limitations are and why fully automated taxonomy or ontology creation will most probably never be possible. See how the resulting semantic knowledge graphs can be used for other purposes, like the extraction of “Shadow Concepts” or graph-based similarities between documents.

Many of those who are involved in content projects have little understanding of the power of taxonomy. You often need to educate business leaders, other UX practitioners, designers, developers, QA, and project managers about how and why taxonomy is useful. They don’t need to create a taxonomy, but they need to be empowered to use taxonomy in their work. What are synonyms good for? How does taxonomy help with the display of content? How can you build a search results page with filters? This talk covers an approach to teaching colleagues about taxonomy and why it should be important to them. Learn how to make taxonomy relevant for a non-taxonomy geek.

Taxonomy and SEO are often run as two independent activities that do not talk to each other. However taxonomy and SEO have a lot in common and can benefit from close collaboration. In particular, managing taxonomy and SEO categories and terms can be coordinated through common practices using a common terminology management application. This presentation discusses some common e-commerce and communications use cases, and describes projects where taxonomy and SEO categories and terms are being managed and published in concert.

Too often, taxonomies are built and launched with little to no thought given to long-term maintenance and growth. After the thrill of the initial taxonomy design and implementation has faded, companies need guidance on how to set up governance committees and work with stakeholders to best control taxonomy growth. This session provides a step-by-step plan to help taxonomists ensure taxonomies continue to provide value for an enterprise year after year.

Dwyer presents a case study of an approach that led to the successful implementation of a taxonomy governance model that was mostly shelf-ware for several years. The governance model was operationalized by IT with heavy business engagement, and roles and responsibilities spread throughout the organization.

Taxonomic work often involves resolving different taxonomies or adapting an established taxonomy to fit custom needs. Ecommerce is a context where there are often multiple competing taxonomies, creating discrepancies between the various ways of organizing product data which can cause confusion among users. Keeping a few key points and best practices in mind can reduce the potential pitfalls of comparative taxonomy development for business applications. Hear about common cross-mapping or resolution challenges (including granularity, terminology versus concepts, and operational constraints) and how to approach balancing external standards and in-house requirements during taxonomy development and maintenance.

In creating customer-facing taxonomy categories, it is important to use vocabulary that users will recognize and place category nodes where they will most likely be anticipated. Where does one begin when faced with stakeholders with competing agendas, distributors that use wildly different terms, and a market that has not yet set a clear precedent? The key is listening to the customer. This talk reviews how analysis of search patterns and seller behavior can shape an ecommerce taxonomy that is customer-focused while still being comprehensible to the distributor/seller/manufacturer.

Best practices in taxonomy design make taxonomies more effective and usable. Hear from a panel of experienced taxonomists about the taxonomies in their organizations and design issues they face, including Indeed’s team process of building a jobseeker-focused multilingual occupational taxonomy (soon to be ontology). Cambridge Healthtech Institute’s SharePoint site for employees across all company departments and for product lines focusing on biotech therapeutics, diagnostics, technologies, and informatics.

The idea of linked data usually generates a lot of interest, but it may be difficult to understand how it will actually benefit taxonomists in their day-to-day activities. How and where will linked data really make a practical difference? Hear about real-life, real-time examples of creating linked data references from external sources such as DBPedia and how we can use that information and append it to data being managed within an existing controlled vocabulary to augment attributes. Sweeney reviews the process of selecting linked data sources, as well as the specific properties from those sources that we wish to append to terms and concepts present within a taxonomy management system.

Many organizations use W3C standards: SKOS for taxonomies and RDF/OWL for ontologies. Until now, there wasn’t a standard for defining the rules for checking that data conforms to these standards, ensuring data quality for consuming applications. Organizations had to either use proprietary approaches, which often come short in supporting requirements. Enter SHACL (Shapes Constraint Language)—a new W3C standard which addresses this problem. SHACL offers rich and flexible notations for expressing practically any rule one can think of. Mekonnen introduces SHACL and the motivations behind its creation and provides several detailed examples of its use to ensure data quality within a vocabulary.

People are at the core of knowledge-sharing—the key to high functioning organizations. In John Seely Brown’s words, “We participate, therefore we are.” New and emerging technology can only enhance learning, sharing, and decision making to create successful organizations. Join our inspiring and knowledgeable speaker as he shares his view of the future of people and tech working together to share knowledge and create winning organizations.

Text analytics and auto-categorization tend to present themselves to the world as esoteric disciplines supported by complex expert systems. Users are immediately confronted by a jargon-wall built with terminology from computational linguistics, such as “tokenization,” “lemmatization,” and “NLP.” At past KMWorld/Taxonomy Bootcamp events, some practitioners who are attempting to get started with auto-categorization projects have voiced a common set of frustrations. Categorizing content shouldn’t require an advanced degree in linguistics. Categorization rules should be simple and transparent. Rules development and taxonomy development should be coextensive rather than separate activities. Rules should be easy to edit, and it should be possible to understand quickly and precisely how changes to taxonomy and rules impact document categorization. This talk explores these issues from a design and user-experience perspective. It outlines a manifesto for demystifying text analytics and for simplifying the process of auto-categorization. The manifesto is aimed at a constituency of content owners and taxonomists and hopes to help them take ownership of the categorization process so they can better control the search and discovery experience for their end users. 

Recently, NPR’s Research Archives & Data Strategy (RAD) team has moved away from their largely manual metadata creation process and incorporated auto-categorization in their digital archiving workflow. In particular over the past year, the RAD team has worked to expand and refine their existing vocabularies to meet the needs of users across NPR and to enhance the accuracy of the auto-categorization. This talk highlights the challenges and lessons learned to date. We discuss the challenges of implementing and fine-tuning the taxonomy management and auto-categorization software and how new tools have had an impact on the workflows of all RAD team members.

Hear about how HBO created its own metadata standard through adaptation of multiple industry metadata standards. Our successful solution is a team effort utilizing taxonomy and ontology expertise, combined with taxonomy data governance, project management, and the right metadata tools. The use of international standards enables the creation of authoritative vocabularies and eases communication between data points. Learn in detail about how to manage taxonomy projects and the type of taxonomy and ontology expertise needed to capture and document the efforts to ensure successful adoption. Hear tips and tricks on how to take existing standards and make them your own, saving time and effort.

Taxonomy development does not scale: companies rarely invest in the human resources needed to build and maintain taxonomies. In 2015, the new enterprise knowledge management team at USAA was tasked with improving the quality of the enterprise knowledgebase. A temporary solution for lack of resources was developed for the pilot, using SMEs with business units as part time taxonomists. The team was so impressed with the results, we expanded the “temporary” solution to our enterprise strategy. Hear how USAA established a new role within the business units to help in optimizing knowledge findability and the taxonomy development process.

Tasked with helping its 2017 Annual Meeting attendees sort more than 200 sessions by personal interest, the American Society of Clinical Oncology (ASCO) leveraged its taxonomy to produce targeted program recommendations. The Session Recommender leveraged existed profile data supplemented with additional information supplied by the user via a short questionnaire to craft individual taxonomy profiles, which were used to generate recommendations. During this session, Travis will offer an overview of the project, including design considerations and challenges, as well as how ASCO intends to use it as the basis for future personalization efforts.

AI and deep learning are hot topics these days, but when you look more closely, most of the advances have to do with image and pat-tern recognition. Taxonomies, on the other hand, have mostly to do with much messier things—words and their multiple meanings. Do these two fields have any common ground? Reamy looks at some surprising ways that companies are learning to take advantage of the strengths of each. With multiple examples from recent projects, he reveals some of the recent successes and roadblocks of combining AI and taxonomy. The next speaker discusses how the business environment is primed for AI and how organizations are learning to incorporate automation in search results as well as in search suggestions. Where would you rank your organization’s competency in automating taxonomy? For most, the answer is simple: It’s not where it needs to be. Historically, taxonomy development has been a manual process. It doesn’t need to be any longer. Hear how three blue chip companies are taking an active approach to the problem, automating the analysis of stakeholder needs and adjusting taxonomy output on-the-fly.

EY’s knowledge architecture team is responsible for developing and maintaining the taxonomies that underpin its core knowledge, service delivery, and collaboration applications. The team currently manages 45 taxonomies, comprising more than 8,000 terms, as well as an auto-classification program. Establishing governance processes is relatively easy; ensuring adherence to them is more challenging, particularly for EY, which employs 230,000 people and operates four core lines of business in 150-plus countries. This presentation describes EY’s overarching taxonomy design, governance model, and taxonomy management processes, including our auto-classification approach. It shares success stories and highlights pitfalls to avoid with a global enterprise taxonomy. Just over a year ago, National Geographic’s taxonomies were only in American English and were U.S.-centric. They now are in 11 languages and counting, a key part of the standard infrastructure tying together National Geographic’s global presence with text analytics support for all languages. This talk is the story of its transformation, and Fulvio shares key lessons and insights gleaned along the way.

Just as humans thrive on nutritious diets, high-performance applications demand well-fed taxonomies and ontologies: rich, up to date, and organically tied to the reality of your data, rather than a simple schema. But spoon-feeding your ontology with facts (or triples) one-by-one is hardly a plan when your knowledge is dispersed throughout a disparate array of information assets. That’s where cognitive technology, and specifically natural language processing, comes into play. Based on your schema, it harvests the instances of your object classes and relationships that actually occur in your content, capturing a wealth of knowledge that makes and keeps your ontology healthy. This session showcases concrete examples of how to successfully implement cognitive technology to feed your ontology and power your information application.

Interested in industry trends? Stymied by a taxonomy design challenge at work? Bring your toughest, crunchiest taxonomy issues and challenges to our panel of seasoned, full-time taxonomists, who will compete to answer your questions with insight, entertainment, and perhaps even controversy! The best questions (as voted by the audience) will bring home prizes!