DrMKZ

CosmoDB Vector searching rather than keyword searching under Azure. Vector searching rather than keyword mashing.

Rather than using keyword, vector converts the searching into mathematical representation.

So you have this Vector similarity, interesting enough we are managing keyword in to numbers. Word embedding software like BERT, Word to Vec, and GPT.

Found this on Medium, graphical representation of word as a searching method,

A Dummy Guide to Word to Vec2. apparently this makes word more efficiently searched.

Word Embedding; changes word into codes. Its like giving a DNA to a word. Specific region of the each letter with their own properties making it into a unique ‘noun’ but in a standardized format.

Exactly like Genetic engineering. Making each word to own its function and attributes. Examples below

Then language modelling, predicting the next word.

Use matrices and vector for words.