Building the Knowledge Graph Behind Direct Answers: A Semantic Ontology Proof of Concept.

About the Client

Company N, a global news and information company, aggregates data from licensed publications, websites, blogs, newspapers, journals, magazines, radio transcripts, images, and newswires, enabling users to mine data, predict market trends, and push relevant content across an organization. Company N’s customer base has long consisted of librarians, researchers, and information managers acting as intermediaries between Company N’s data and the business users who ultimately need answers, not article lists.

Industry Context

By 2010 and 2011, semantic web technologies — ontologies, RDF triples, SPARQL, and knowledge graphs — were still genuinely early in enterprise adoption; Google’s own Knowledge Graph, built in part on Freebase, which Google acquired in 2010, would not launch publicly until 2012. For an information intermediary business like Company N, most information products up to that point were fundamentally search-and-retrieve, returning a ranked list of articles for a trained researcher to synthesize into an answer. Moving from returning articles that might answer a question to directly answering the question itself, using facts extracted and linked across content, represented a genuinely disruptive shift — one significant enough, given how unproven the technology still was, to warrant validating through a disciplined proof of concept before any larger product investment.

Current Situation

The output of a Company N search includes a list of articles and data that potentially answer a business user’s query, and a librarian must then synthesize that content into a research product meeting the user’s actual requirements. Company N’s working hypothesis was that applying more detailed content descriptions, linking concepts and facts in a graph database, and building reports that directly answer user questions would provide the foundation for a portfolio of high-value products, and it engaged Iknow to test that hypothesis.

• Article lists instead of direct answers. Company N’s core product returned lists of articles rather than direct answers, requiring a human librarian to synthesize the actual answer to a business user’s question.
• No proven approach to semantic content linking. No proven approach existed for using ontologies and semantic technologies to link concepts and facts across Company N’s vast content base.
• Immature, unproven underlying technology. The underlying semantic web technologies were still genuinely early in their lifecycle, with limited enterprise precedent for applications like this one.
• No evidence base for a larger product investment. Company N needed evidence, not assumptions, that a semantic approach could deliver genuine value before committing to full product development.

• Assess the feasibility and utility of applying ontologies to Company N’s content.
• Recommend the technology stack and architecture for describing content with an ontology and extracting structured facts.
• Develop a proof of concept to test the real-world development challenge and cost-benefit of the approach.
• Recommend a new product strategy and roadmap based on the proof of concept’s results.

How Iknow Structured the Work

Iknow structured the engagement across four sequential phases — feasibility assessment, technology stack recommendations, proof of concept development, and product strategy recommendations — reflecting Iknow’s knowledge graph and ontology design methodology for validating an unproven technology approach with real evidence before committing to full-scale product development.

Key Activities & Decisions

• Phase 1: Feasibility assessment. In Phase 1, Iknow assessed the feasibility and utility of ontologies for adding value to Company N's content, delivering a report and presentation on how ontologies work and the feasibility of applying one to Company N’s content.
• Phase 2: Technology stack recommendations. In Phase 2, Iknow recommended a stack of technologies and an architecture required to describe content with an ontology and extract triples in outputs that met client requirements.
• Phase 3: Proof of concept development. In Phase 3, Iknow developed a proof of concept to test the magnitude of the development challenge and the cost-benefit of describing content with an ontology.
• Phase 4: Product strategy recommendations. In Phase 4, Iknow recommended a new product development strategy and next steps to monetize the solution, based on the proof of concept’s results.
• Semantic enrichment and user interaction architecture. Iknow designed a semantic enrichment and user interaction architecture, combining industry-specific knowledge modules and linguistic rules with a knowledge extraction engine that feeds a federated knowledge base, alongside a user interaction path that translates keyword queries into structured SPARQL queries.
• Oil and gas proof of concept. Iknow collaborated with one of Company N’s oil and gas clients to develop a proprietary proof of concept in oil field acquisitions, linking descriptive oil sector data, buyer and seller information, location details, valuations, and transaction prices.

Stakeholders & Collaboration

Iknow served as prime contractor, working directly with Company N Company N and partnering with one of Company N’s own oil and gas clients to design, build, and evaluate the proof of concept.

Testing an Unproven Technology Without Overcommitting Resources

Ontologies, RDF, and SPARQL were still maturing technologies with limited enterprise precedent, and committing to full product development before establishing genuine feasibility risked significant wasted investment. Iknow addressed this by structuring the engagement into four disciplined phases, moving from feasibility assessment through technology recommendations to a scoped proof of concept before any product development commitment was made.

Proving Genuine Business Value in a Credible, Real-World Domain

A generic or hypothetical demonstration risked proving technical feasibility without proving the approach could support an actual product. Iknow addressed this by partnering directly with one of Company N’s real oil and gas clients to build the proof of concept around oil field acquisitions, a genuinely valuable use case, rather than an artificial demonstration.

Quantitative Outcomes

• Project structure: A four-phase project delivered over nine months, from feasibility through product strategy.
• Architecture delivered: A complete semantic enrichment and user interaction architecture designed, including a federated knowledge base and SPARQL-based query capability.
• Proof of concept completed: A successful, working proof of concept built with a real Company N oil and gas client.
• Roadmap delivered: A 10-year, phased ontology implementation roadmap.
• Engagement duration: Nine-month assignment.

Qualitative Outcomes

The project resulted in a successful proof of concept and a new product development timeline. Because semantic technologies were in their infancy, Iknow developed a rather long 10-year implementation roadmap, giving Company N time to refine and iterate on its approach — a deliberately paced, realistic path rather than an overpromised, compressed timeline. The oil and gas proof of concept gave Company N tangible evidence, grounded in one of its own client relationships, that the semantic approach could work and could ultimately support higher-value products beyond simple article retrieval.

Timeline to Impact

During the nine-month engagement, Iknow delivered the full feasibility assessment, technology recommendations, a working proof of concept, and product strategy, giving Company N a structured path from initial product delivery to fully realized semantic products within four years.

Core Skills

• Knowledge graph design and implementation
• Ontology design and development
• AI and semantic technology proof-of-concept development
• New product strategy for information services

Methods & Frameworks

• Four-phase feasibility-to-product-strategy methodology
• Services-oriented semantic enrichment architecture
• Client-partnered proof-of-concept validation

Technologies & Tools

• Ontologies, RDF triples, and SPARQL query technology
• Federated knowledge base and triple store architecture