The Digital R&D Division of a global manufacturer of home care and household cleaning products wanted to have access to reliable, complete and unambiguous information about surfactants and reported skin interactions for their research and development (R&D).
When developing cleaning agents and personal care products, the main concerns are related to negative skin reactions from exposure to certain ingredients. On top of that, there is a mounting pressure from consumers to use natural ingredients instead of chemicals. All this generates a significant amount of research before the release of any new product.
To make things even more complicated, the entries for natural surfactants and other complex mixtures of components in public sources such as PubChem or ChemSpider are incomplete and often inaccurate. The existing models and ontologies often don’t include the concepts required to describe such materials.
So, the Digital R&D Division was frequently faced with ambiguous, fragmented and insufficient data, which they had to process manually to identify potential risks. Consequently, their product R&D took a lot of time and effort, and could easily miss possible relationships between a surfactant and a negative reaction reported in the scientific literature.
As the company put very high priority on identifying the right candidates for their products, they wanted to create a comprehensive knowledge graph (KG) with information about surfactants and their relationships to skin interactions.
Due to the very little or no coverage of information about surfactants in the existing domain ontologies, the main challenges in this project were:
Ontotext’s first task was to create a brand-new taxonomy for substances, compounds and mixtures classified as different types of surfactants.
Ontotext was also provided with a set of PubMed scientific articles about surfactants and reported skin reactions (redness, dryness, itchiness, blistering, inflammation, etc.). Although limited, this set was selected by the company to sufficiently cover the diversity of document types and associated challenges.
Using the newly created Surfactant Ontology, Ontotext processed the trial set of articles, applying text analysis techniques such as Name Entity Recognition, rule-based extraction and relation extraction of specific properties and values. All the extracted information was used to create a domain-specific knowledge graph. By representing the concepts and their relationships through formal semantics, and by enabling automated reasoning, the knowledge graph allowed the R&D Division to uncover unobvious / hidden correlations between surfactants and reported issues in specific documents.
Ontotext’s unique approach is based on the solid integration between knowledge graphs and text analysis. The information extracted by text analysis enriches the knowledge graph and the ontology, and this, in turn, enables richer findings. The result is an advanced intelligence tool for conducting quick and efficient product research in large volumes of unstructured data. The flexible and dynamic model of the knowledge graph allows the resulting valuable knowledge to be constantly enriched with new information.
Ontotext’s approach also allows fast onboarding of new employees who can immediately dive into the domain-specific knowledge, which was previously locked in documents and in the heads of individual researchers.
Ontotext’s solution was built for a very specific potential health problem, but the functionality is applicable to all types of Life-sciences, Pharma and Healthcare domains as it is based on a generic technology.
Do you think this case resembles your particular needs?