Waylay

-D’Haeseleer Waylay is a low-code platform that enables developers to benefit from enterprise level automation. By building small code snippets, or by using existing ones, users can chain together logical operators to create automated rules – such as turning on the water sprinklers when there has been sunny weather without rain for 3 days, or scheduling an inspection of an industrial machine if its sensors detect any anomalies. By making this technology accessible to everyone Waylay’s core values are fulfilled. For example, imagine being able to communicate with machines over voice and text control in a natural manner; instead of having to interact with a computer through traditional methods. This is where Natural Language Processing (NLP) comes into play. However, creating deep learning solutions based on these traditional methods faces certain challenges – primarily due data scarcity and the need for retraining every time new cases arise. Prompt engineering provides us with a solution: leveraging GPT3 allows us to build highly data efficient systems which don’t require retraining each time different types of Waylay rules occur or different manners of speaking have to be taken into consideration. The idea behind this is that we let GPT3 output canonical sentences so that our natural language input can be structured more clearly and efficiently translated into something Waylay understands better. In summary, by translating our semantic parsing task into a translation task we have managed to take advantage of large pre-trained language models such as GPT3 in order create complex automation software without requiring much data points nor needing frequent retrains – even across languages!