From Broad Terms to Precise Answers: Rebuilding Company Z Manufacturing’s Search Taxonomy.

About the Client

Company Z is a global pharmaceutical manufacturer. Its manufacturing division oversees the formulation, packaging, and distribution of its global product portfolio across an interdependent manufacturing network. The manufacturing division’s Knowledge Management Center of Excellence maintains 21 active communities of practice that curate manufacturing-related technical documents and host discussion forums, enabling scientists and engineers across its global facilities to post questions, answer them when they have relevant expertise, and read others’ answers — a capability that promotes rapid problem-solving across Company Z’s manufacturing sites worldwide.

Industry Context

Communities of practice have become a standard knowledge management technique in life sciences manufacturing because so much practical expertise resides in the minds of scientists and engineers rather than in formal documentation, and forums like Company Z’s give that tacit knowledge a searchable home. But a community’s value depends entirely on whether people can find what has already been discussed, which in turn depends on the taxonomy underlying the search experience. Iknow has been a partner of Progress Software, the maker of the Semaphore content intelligence platform, since 2009, giving it deep familiarity with Semaphore’s model-driven approach to autoclassification — building rule sets from a curated taxonomy rather than using a machine-learning approach that requires constant retraining as new content appears. That distinction mattered directly to Company Z, whose technical vocabulary evolves as quickly as the science itself.

Current Situation

By 2020, the roughly 400-term legacy taxonomy used to tag Company Z’s 21 communities of practice was too broad to describe their technical content, and users were complaining about imprecise search results. Company Z’s KM COE asked Iknow to mine the community SharePoint sites and discussion forums for terms that more accurately described the content, then use those terms to expand the enterprise taxonomy, add synonyms and related terms, and update Semaphore’s autoclassification rules — all toward a single enhanced taxonomy capable of supporting every Company Z search application.

• A taxonomy too broad for genuinely technical content. Company Z’s roughly 400-term legacy taxonomy could not capture the granularity of its diverse technical domains — analytical sciences, biologics, vaccines, sterile processing, cleaning, and automation — leading to imprecise search results across all 21 communities.
• Current domain knowledge is trapped in unstructured discussion threads. The precise, up-to-date terminology Company Z needed already existed in years of community Q&A discussions, but it had never been extracted and formalized into the taxonomy.
• Enterprise-wide ripple effects. Any change to Company Z’s shared primary taxonomy and autoclassification rules would affect search across Company Z’s entire system, not just a single community. Therefore, changes had to be carefully validated before being released to production.
• Twenty-one distinct technical communities to satisfy. Getting new terms, synonyms, and taxonomy branches right required genuine subject-matter validation across a diverse set of disciplines, not just the central KM team’s best judgment.

• Mine Company Z’s community SharePoint sites and discussion forums for terms that more accurately describe the technical content.
• Expand and update Company Z’s enterprise taxonomy with new terms, synonyms, related terms, and new taxonomy branches where warranted.
• Update Progress Semaphore’s autoclassification rules to improve the accuracy and precision of metadata tags.
• Deliver a single enhanced taxonomy that supports all of Company Z’s search applications and systems.

How Iknow Structured the Work

Iknow structured the four-month engagement into two phases: an analysis and mapping phase that proposed taxonomy changes, followed by a validation and testing phase that confirmed those changes with community SMEs before moving anything into production.

Key Activities & Decisions

• Taxonomy analysis. Iknow analyzed terms in the communities’ existing taxonomy for redundancies and gaps, then mapped those terms to Company Z’s primary enterprise taxonomy.
• Proposed modifications. Iknow proposed modifications to the primary taxonomy and compiled a structured list of questions for the communities’ SMEs covering proposed changes, synonyms, acronyms, and other concept labels.
• Autoclassification impact review. Iknow reviewed the existing autoclassification process for the community sites to understand how proposed rule changes would ripple through production search.
• SME interviews. Iknow interviewed selected community SMEs to understand their search requirements and validate proposed terminology.
• Finalized taxonomy. Iknow finalized the mapping and additions to the full enterprise taxonomy, the deletions from the community-level taxonomy, and any corresponding tag reassignments.
• Rule updates and testing. Iknow updated and tested the new Semaphore classification rules against real community content, then developed recommendations for future search integration efforts.
• Production quality control. As a final step, Iknow tested the updated taxonomy and rules on new content before releasing the enhanced taxonomy into production.

Stakeholders & Collaboration

Iknow served as prime contractor, working directly with Company Z’s Knowledge Management Center of Excellence and subject-matter experts across the 21 communities of practice.

Expanding the Taxonomy Without Fragmenting It

Adding terminology from 21 diverse technical communities risked pulling apart Company Z’s enterprise taxonomy into inconsistent, community-specific vocabularies. Iknow addressed this by first mapping each community’s terms against Company Z’s primary taxonomy to identify genuine redundancies and gaps, so every addition strengthened a coherent enterprise structure rather than fragmenting it further.

Validating Scale With Real Subject-Matter Expertise

Text mining alone could surface candidate terms, but only genuine domain experts could confirm which were accurate, which were synonyms of existing terms, and which reflected real gaps. Iknow addressed this by compiling structured SME question sets and interviewing community experts directly, ultimately validating more than 1,000 new terms with the people who used that vocabulary every day, rather than relying on text-mining output alone.

Quantitative Outcomes

• New terms validated: 1,000+ new taxonomy terms mined from community content and confirmed with subject-matter experts.
• Communities covered: 21 active communities of practice spanning Company Z’s manufacturing science and engineering disciplines.
• Legacy taxonomy modernized: the roughly 400-term legacy taxonomy was analyzed, expanded, and integrated into Company Z’s primary enterprise taxonomy.
• Delivery timeline: completed in four months.

Qualitative Outcomes

Company Z achieved a step-function improvement in the quality of its search results. Because the new terms captured the communities’ current discussions rather than a static reference list, the enhanced taxonomy reflected the cutting edge of pharmaceutical manufacturing science and engineering. The new terms were added to the Progress Semaphore ontology to autoclassify both the questions and answers in Company Z’s community forums and the technical documents its communities curate, extending the benefit across Company Z’s search ecosystem rather than to a single site. The engagement also marked a deepening relationship: having earlier delivered a taxonomy strategy workshop for Company Z, Iknow returned as prime contractor for full implementation — a sign of the trust Company placed in its taxonomy expertise.

Timeline to Impact

Iknow delivered both phases of the engagement — analysis and mapping, then SME validation, rule updates, and production testing — within a four-month period, taking the taxonomy from initial mining through to production-ready release in a single engagement.

Core Skills

• Taxonomy development & governance
• Content mining & text analytics
• Semantic autoclassification / ontology management
• SME-driven validation methodology

Methods & Frameworks

• Two-phase mine-map-validate methodology
• Structured SME interviews and question design
• Taxonomy mapping and gap analysis
• Production quality-control testing

Technologies & Tools

• Smartlogic Semaphore (Ontology Manager, Classification & Text Mining Server)
• SharePoint community sites and discussion forums