Alexa for coatings laboratories

Voice-controlled digital assistants have become a part of many people's daily lives. In the future, they will also make laboratory work easier. The first such assistant is now really

.
names like Siri and Alexa, the virtual assistant is now a family member in one in five American families. The Voice Control Digital Assistant is committed to helping us in our daily lives. If it all goes according to Dr. Gaetano Branda's wishes, it will be just the beginning. As head of Ingtrove's coatings additives business, Branda wants to transform voice-controlled digital assistants from simple assistants in everyday life to chemical experts and use their language where comprehensive expertise and technology are needed: laboratories.
in order to fully meet the customer's color, gloss and durability requirements, experts must create complex mixtures in the laboratory and supplement them with the right additives. Thousands of combinations are possible - actually far more than the human brain can handle.
"the usual assistants simply don't understand our language," says Dr. Oliver Kröhl, head of strategic business area development for Ingtrove Coating Additives.
formulater spends considerable time searching for notes and data sheets. Therefore, the new digital assistant Coatino can help users study adjusting ingredients directly in the lab. "We talked about new ways of growing our business," says Dr. Oliver Kröhl, Head of Strategic Business Area Development and Project Manager for Coating Additives. Instead, you need to prove that you can come up with solutions in the form of new services and business models. The
from cans to prototypes
to determine whether the digital assistant worked properly, the scientists painted an empty can of paint with the company's colors. They then put it in the lab, where they filmed a discussion between colleagues and cans. In the video, the user asks the can for a water-based desiccant for wood coatings. The cans gave a reply, provided a range of products to laboratory staff, and ordered samples.
"At the time, the question was answered by a colleague standing behind a wall," says Kröhl. "The video also had structured interviews with several people sharing customers and teams.
this approval encourages developers to enter uncharted territory. "We're experts in paint and coatings, but not voice control assistants," says Kröhl, "which is why we know this project may not work." However, we and our customers believe that it has such great potential that we are willing to take risks.
learning paint language
is not easy because traditional speech recognition systems cannot handle professional vocabulary. "The usual assistants simply don't understand our language, " says Kröhl. For example, when you ask them about dispersion, fluidology, or silicone resins, they quickly reach their limits, and they can only provide general information at most.
"They have to be able to do more to make the coating, " says Kröhl. "If they don't understand the properties of the components and how they interact, they won't help in the lab." There
thousands of combinations, depending on
, the various ingredients affect each other's effects. The number of possible combinations is huge. Even if only ten curing agents, ten adhesives, ten pigments and ten additives are considered in the development of coating formulations, these numbers can be translated into 10,000 possible combinations. "Customers have very precise ideas about what the product should have once it's finished," says Blanda.
to develop a functional voice control assistant for the coatings industry, researchers first began building all available information and input it into a large database. In the next step, they can use voice control to call up this information.
global application training
for example, if you ask the assistant, "Which additive is suitable for printing inks?" "It is clear that the system must be able to understand each word. In addition,
"assistants can tell me which additive is best for my formulation and my requirements," explains Dr. Gaetano Blanda, head of Ingtrove's coatings additives business.
Coatino must understand that "additives" specify a certain category of coating components. In the next step, the assistant must access its data, search for it, create the appropriate link, and assign the data to possible related results. , which first breaks down the sound sequence into the smallest components and searches for data based on feature properties.
particular challenge for the assistant is that it must be able to understand not only the German nouns in the main grid, but also the German nouns in other cases. The researchers also want to make sure that the speaker's dialect or accent doesn't interfere with the outcome. The aim is to enable Coatino to understand the pronunciation of customers around the world. In addition to these challenges, there are different conversation speeds and rhythms of the speakers and the specific context of the discussion.
"The training process is very troubling, " says Kröhl. " "For nearly two years, Coatino has been jointly developed and trained by the Berlin line of business and external development companies. The assistant presented the prototype at the 2019 European Coatings Show in Nuremberg in April and passed its first major development test.
when asked about suitable additives, the assistant not only provides a list of products, but also prioritizes them. "Coatino can tell me which additive is best for my formula and my requirements. It can provide me with available advice," Blanda said. Once the user has found the product he or she needs, he can issue voice commands to order samples, call the relevant technical data sheet directly via e-mail, or have a conversation with the scheduled expert.
a new recipe from the
The Coatino prototype was ready in time for the start of the European Coatings Show. "We immediately presented it to some of our customers," Blanda said. "Users no longer use cans, but pass their wishes to the tablet via a microphone. In 2020, researchers plan to use Coatino throughout the coatings industry.
, the further development of the system is not over. "When you use a digital assistant, you keep thinking about new features," says Kröhl. For example, Coatino can imagine not only offering existing formulations, but also suggesting its own new mixtures. "Our Coatino could one day really be an artificial intelligence entity," said Branda, who scientists can test the mixture directly in the lab. "But we still have a long way to go."