Knowledge Graphs for Autonomous Formulation

"Autonomous Formulation" refers to the use of the autonomous software or robots to determine the composition and recipe of mass-market consumer products. The robots used to carry out these experiments generate a vast amount data along with varying quantities and qualities of metadata. Exploring this data is of high importance to the material scientists as it can contain information regarding complex behaviours and characteristics.
We proposed the introduction of knowledge graphs to enhance the quality of information that can be extracted from the data. Knowledge graphs, as popularised by Google in 2012, are a collection of facts, or triples, that are constructed with two concepts, or classes, and the relationship between them. As these relationships are defined by an ontology, the knowledge within a knowledge graph can be reasoned upon.
Within the use case of experimental data generated by robots, the structure provided by storing this data in a knowledge graph can allow material scientists to identify more complex relationships and behaviours within the data. This can involve activities such as investigating and predicting the effect that changing different experimental parameters may have on the results of a given experiment. Insights of this nature can help increase the speed of development of products as well as uncover previously unknown pathways of research.
This work aims to develop and implement a process to (semi)automatically generate a knowledge graph for data collected by robots operating in a laboratory setting, with the goal of assisting material scientists enhance their product development pipeline.