Adding an AI-based semantic ticket search

In this section, we’ll add a semantic AI search to our tickets, such that there’s no need for the search to match exact words anymore but simply to match with the meanings of words. To implement an AI-based content search, we store a so-called embedding, a mathematical vector representation of the content alongside the row.

Even though embeddings are far more complex than this, I’d like to give you an analogy to understand embeddings at their core. Think of two topics, dogs and cars. Let’s say, for a given set of words, we define the amount of dog topics and car topics between 0 and 1, where 1 means “highest possible semantic meaning of topic” and 0 means “no semantic meaning of the topic.”

Now let’s take the sample phrase, “My dog likes going for a walk.” How much “dog” is in this phrase semantically? There’s no clear answer to this as this...