Case study How Cortical.io's semantic search engine is helping a global company automate its marketing claim verifications

Customer

Consumer goods giant

Goal

Automate the task of finding research studies to support marketing claims

Challenges
  • Ensuring that marketing claims about a product are not misleading or deceptive is a major challenge for companies that must comply with consumer protection laws and industry self-regulatory guidelines
  • Marketing claims are most trustworthy when supported by sound or factual research results, for example from clinical studies and consumer reports
  • Almost 300,000 new clinical studies are registered every year, making it impossible to manually keep track of new evidence related to a specific product
  • The scientific vocabulary used in such studies strongly differs from the marketing jargon used to describe a product to consumers
  • The product-relevant scientific information is often not formulated in words, but encapsulated within chemical formulation
  • As a result, the verification of marketing claims is a long and tedious process involving human-based tasks of searching, and manually validating evidence documentation
Solution
  • Unsupervised training of Cortical.io’s semantic search engine to learn the vocabulary used in the company’s marketing claims and evidence documentation (consumer reports, clinical study reports, fact sheets, other scientific material)
  • Configuration of a data extraction engine for common names, formulation and ingredients of consumer products
  • Enabling the filtering of search results based on common formula name, formula coding, or a combination of formula ingredients
  • Enabling searching with queries formulated as marketing claims to identify related scientific information
Results
  • Discovery of new sources of evidence within scientific literature
  • Automation of the marketing claim verification process
  • Reassignment of claims support research to junior researchers
  • Reallocation of expensive subject-matter experts’ time to more demanding tasks
The Cortical.io difference
  • Cortical.io’s semantic search engine understands the meaning of marketing claims and matches them with research studies, even if they do not use the same vocabulary
  • Extraction models can be trained very quickly, based on annotations by the firm’s subject-matter experts on only a handful of sample clinical reports
  • The solution instantly extracts quality structured data from research studies and other sources of evidence
  • The feedback of subject-matter experts on the quality of search results is directly looped back into the system to facilitate continuous learning while preserving consistent results