Which data field is the vector crated from

Short Course Q&A Vector Databases: from Embeddings to Applications
1

In the chapter “Vector Databases”, it wasn’t very clear as to which specific data field was sent to the embedding model to create the vector. We have 3 fields - Category, Question, and Answer. Is the vector generated from Question or Category or Answer field of the data?
One of the headings in the notebook says " Let’s Extract the vector that represents each question!" which means the vector was based on the Question field, but where was this defined in the code? How would weaviate know which field to create the vector from?

2

Hi @llm_is_my_world,
Welcome to the community :smile:

You can find the answer to the question here. In this official blog by Weaviate, they have decoded the functionality of the text2vec module.

In short, from what I can understand from this blog, the following is used to create the vector:

answer {corresponding value of answer} category {corresponding value of category} question {corresponding value of question}

which is essentially the concatenation of all the test-based fields, further converted to lowercase.

Let me know if this helps.

Cheers,
Elemento