Developed to Support STM Product Deployment
Cortex is also designed to fill the gap between unwieldy academic taxonomies (e.g., SNOMED) that produce hurdles for implementation in active applications and sparse production taxonomies (e.g., MeSH, which contains about half the number of concepts and equivalents currently available in Cortex) that do not cover concepts with enough detail. Cortex is efficiently structured to enhance the speed of software request processing and quick and straightforward application to content by automated routines, while at the same time being complete enough to cover topics at appropriate levels of granularity.
Connection to Real-World Users
Cortex reflects a unique view into new terminology being used by practitioners and researchers both formally and informally, as used in “real world” situations in hospitals, outpatient clinics, research labs, and academic settings. Domain experts and editorial boards provide input on concepts and hierarchies that best support the varied use models for our applications.
Agile and Up-to-Date
An effective taxonomy is a living data structure. Taxonomies are not one-time development projects—they must be maintained on an ongoing basis as a domain and the demands of its users evolve. With the pace of research today, taxonomies without a continuous maintenance plan will be outdated before they can even be deployed. Silverchair staff update the Cortex taxonomy daily in response to new topics found in breaking news, journals, and books. Our focus is capturing evolving concepts as they appear in the literature—and when they are of heightened interest to information seekers. Continuous, rapid contributions from the STM literature, coupled with a rigorous management process provided by Silverchair, produce the most current, actionable taxonomy available in science, technology, and medicine.
In addition to analysis of emerging literature, Cortex is expanded through analysis of search logs reflecting millions of searches per month and user behavior across dozens of information sites to learn how users look for topics. This activity supports continuous addition of jargon, “common parlance,” alternate search phrasings, abbreviations, and even common misspellings to the database of synonyms and other lexical equivalents used in conjunction with Cortex. For example, equivalents for “myocardial infarction” include heart attack, MI, and myocardial infraction.
Publisher-specific taxonomies developed at the enterprise level are often ineffective at connecting your content to other resources in the field. Cortex solves this problem by delivering domain-level concepts that are, whenever possible, mapped via Concept Unique Identifier to the Unified Medical Language System (UMLS, which includes MeSH, SNOMED, ICD, RxNorm, and many other standard health care taxonomies) developed and maintained by the National Library of Medicine. By maintaining this connection to UMLS, Cortex ensures high levels of interoperability with UMLS constituent vocabularies, which in turn offers high levels of interoperability for health care partnerships such as with electronic health record providers.
Automated but Human-Tuned
Silverchair’s semantic tagging process is automated, allowing publishers to feed large volumes of content through the system rapidly and with little incremental cost. However, that automation is informed by in-depth content analysis and a project-specific configuration process performed by Silverchair semantic tagging experts. The outcome of this process is the careful tuning of TagMaster to ensure accurate semantic tagging for the content format, topic areas, and use case. The autotagging configuration for a research journal is different from the configuration for a textbook. A publication aimed at clinicians will be tuned in a different way from one geared toward bench scientists. And a product designed for quick reference will have a different set up from one designed for long-form reading.
Granular and Summary-Level Tagging
Different use cases require different levels granularity in semantic enrichment. Placing journal articles in relevant topic collections, for example, requires tagging at the article level. The reader expects that the article as a whole is largely relevant to the collection. On the other hand, building an image library or linking to the relevant table in a quick-reference application requires tagging at the image, table, and paragraph level. TagMaster applies tags at the most granular levels (paragraph, images) and considers the significance of each tag applied to calculate summary tags not only for the chapter or article overall, but for each section within that chapter or article. This allows content to be rapidly deployed as new use cases and product features emerge.
Your semantic tagging strategy should be guided by the needs of the users of your content and your information products. A clinician wants to know what’s the best treatment for the patient she’s about to see. The researcher wants to know everything about a subject of interest. The student wants to prepare for an exam. Their use cases—and their demands on your product—will be different. Tagmaster was created by Silverchair to allow content tagging that can support a variety of use cases—often at the same time. The ability to tune your semantic metadata to fit particular use cases allows increased levels of site sophistication and personalization, better user experiences, and a more flexible product development architecture.