Soda to milk: Electronic AI tongue detects liquid samples with 80% accuracy

Researchers at Penn State have developed an extraordinary new tool: an electronic tongue capable of discerning subtle differences in liquids.

Notably, this innovative device can distinguish between various types of milk (even with water content), sodas, and coffee blends. Interestingly, it could even detect signs of spoilage in fruit juices and could be used in quality control.

The researchers used artificial intelligence to interpret the data generated by the electronic tongue, resulting in even greater accuracy.

Tongue equipped with sensor

The gustatory cortex is a part of the brain that processes and understands tastes. It goes beyond the basic five tastes (sweet, sour, bitter, salty, and savory) to recognize more complex flavors.

The researchers created a neural network to imitate how the gustatory cortex works. This neural network is a machine learning algorithm that can learn and understand data like the human brain .

The electronic tongue uses a “graphene-based ion-sensitive field-effect transistor” or a sensor to detect chemical ions. This sensor is connected to an artificial neural network that has been trained on different data sets.

Interestingly, the sensors used in the electronic tongue are versatile and capable of detecting various chemicals without requiring specific sensors for each type.

Researchers trained the AI using 20 parameters related to how the liquid interacts with the sensor.

Remarkably, the AI accurately identified different samples, including milk, soda, coffee, and fruit juice, with over 80% accuracy within minutes.

“After achieving a reasonable accuracy with human-selected parameters, we decided to let the neural network define its own figures of merit by providing it with the raw sensor data. We found that the neural network reached a near ideal inference accuracy of more than 95% when utilizing the machine-derived figures of merit rather than the ones provided by humans,” said Andrew Pannone, co-author and a doctoral student in engineering science and mechanics.

AI’s decision-making

After the initial results, the researchers used “Shapley additive explanations” to understand how the neural network made decisions.

“This approach uses game theory, a decision-making process that considers the choices of others to predict the outcome of a single participant, to assign values to the data under consideration,” the press release explained.

It was found that the neural network considered multiple factors together, rather than just individual human-assigned parameters.

As per the press release, this method provides insight into the neural network’s decision-making process, which has been difficult to understand in the field of AI.

“We found that the network looked at more subtle characteristics in the data — things we, as humans, struggle to define properly,” said Saptarshi Das, corresponding author.

Das explained: “And because the neural network considers the sensor characteristics holistically, it mitigates variations that might occur day-to-day. In terms of the milk, the neural network can determine the varying water content of the milk and, in that context, determine if any indicators of degradation are meaningful enough to be considered a food safety issue.”

The electronic tongue’s capabilities depend on the data it is trained on. In addition to liquid assessment, this technology could also be used for medical diagnostic purposes.

The findings were published in the journal Nature.