Vectors represent unstructured data and are essentially lists of decimal numbers. When vectors are used in semantic search, we refer to vector embeddings.
the vector embeds the structures and relationships of the original data (which may be the semantic meaning of a paragraph or the attributes like colors, angles, lines, and shapes in a picture, and so on).
A vector is a mathematical structure with a size and a direction. For example, we can think of the vector as a point in space, with the “direction” being an arrow from (0,0,0) to that point in the vector space.
As developers, it might be easier to think of a vector as an array containing numerical values. For example:
When we look at a bunch of vectors in one space, we can say that some are closer to one another, while others are far apart. Some vectors can seem to cluster together, while others could be sparsely distributed in the space.
The availability of pre-trained machine learning models has helped spread and standardize the approach. In practice, many machine learning projects use pre-trained embedding models as a starting point, and benchmarking and fine-tuning them for the specific task helps introduce semantic search into a service
When a machine learning model generates vectors, they embed the distinctive features of data into floating point numbers of fixed size in a compact and dense representation and translate the human-perceived semantic similarity to the vector space.
Text Embeddings:
Image embedding:
If we want to compare 2 sentences only using vectors and the proximity between them we just compare the words accuracy and not compare the semantic meanings of the sentences.
So, to solve that problem, we can use an embedding model like mxbai-embed-large, This types of model are trained with amount of labeled data to a neural network.