I am new to this technology and trying to understand the applicability of taking a large time series based dataset largely made up on metrics and associated meta data and storing the embeddings in the pinecone vector database. The aim is to use Question Answer to allow users through a ChatGPT style LLM ask investigatory style questions of the measurement data. Conceptually thinking of leverage the context of the user asking the question to apply filters against the meta data stored in the time series data. Thanks for any pointers!!
Hi @camte,
Can you give an example of the time series data you’re working with? How your source data is structured will have a large impact on both what model you’re best off using to convert it into vectors, and what kind of semantic search your application will be capable of performing against it.
Thanks for your quick reply @Cory_Pinecone. Our data set is largely measurement metrics that are relevant to the meta data of each sample. You could think of it like health metrics such as pulse rate, blood pressure etc and the meta data being the person being tested, their age, location etc etc. Our aim is to be able to ask questions across a large collective dataset of these ongoing measurements to uncover insights that may otherwise be lost in standard visualisations etc
Hi @camte , have you got any updates on this question?