When building a RAG system, what is the risk of making your document 'chunks' too large before embedding them?

If a chunk contains too much information, its vector representation becomes 'diluted' or fuzzy, making it harder for the database to find a specific match for a narrow user question.

When building a chatbot using a vector database, what happens to the user's initial question?

To find a relevant answer, the system must convert the query into a vector (embedding) so it can compare the mathematical distance between the query and the stored knowledge data.

In the context of word embeddings, what does the 'distance' between two vectors primarily represent?

Embeddings use mathematical distance to represent how related two entities are. While common letters or dictionary order are attributes of words, they do not define the vector's position in an embedding space.

Learn Vector Databases - Free Learning Path