Formulation, process, performance — in the field of new materials these three kinds of data are never independent. Formulations carry conventions, process and performance are tightly coupled, and performance is incomparable once its conditions are missing. The hard part of data work in this domain is not pulling the numbers out, but making sure the numbers pulled out can still be compared.
Component content can be expressed as wt%, mol%, at% or even volume ratios, and trace components are often glossed over as "balance" or as range values. These notations are not sloppiness — they are the applicant's definition of the boundaries of the formulation. Convert them into a unified number and the information is lost.
Conversion itself has preconditions: going from mass fraction to mole fraction requires the molecular weight of every component, which patents do not necessarily provide. Without that precondition, any conversion is an untraceable estimate.
Sintering temperature, pressing pressure, annealing time, atmosphere conditions — a change in any one of them alters the final performance. Extract only the performance values and not the corresponding process, and this batch of data cannot answer "why is this formulation better".
Whether particle size is D50 or D90, whether relative density is based on theoretical or measured values — each patent writes it differently. A data point with an unclear convention is not a sample once it enters the database; it is noise.
Formulation, process and performance are split into three mutually linkable tables, chained together through publication numbers and example numbers, so that you can query a single category of data on its own and also reconstruct the complete record of "one set of formulations + one set of process steps + one set of performance data".
Turn the formulations, process steps, characterization data and performance metrics in patents into a comparable database. Keeps the original wt% / mol% conventions and test conditions, without forcing conversions or turning range values into point values.
Ionic conductivityProcess parametersFormulation conventionsNo. The original conventions are kept as written and annotated separately. Converting between the two requires the molecular weight of every component, which patents do not necessarily provide — forcing the conversion introduces untraceable error, so keeping the convention is actually the more rigorous choice.
Because a value divorced from its conditions is not comparable. Take ionic conductivity: values at different temperatures can differ by several orders of magnitude; without the test temperature the number, though real, cannot be placed alongside another dataset for comparison.
They are chained together through publication numbers and example numbers, so that one record carries a set of formulations, a set of process steps and a set of performance data at the same time. Stored separately, the three lose causality — only together can they answer "why is this formulation better".
The convention is recorded as written in the source (for example, whether particle size is D50 or D90, or whether relative density is based on theoretical or measured values), without conversion or omission. A data point with an unclear convention becomes noise rather than a sample once it enters the database.
Send us a representative patent family and we can run a small trial annotation first, so you can see directly how formulation conventions and test conditions are handled.