Hi, i am quite comfortable with the idea of a billion node graph ( ala social media site ) where most the the data can be neatly ( relatively ) encapsulated as objects with properties. What i am unable to even visualize is how this can incorporate , lets say, sales data for a product line by time and geographical granularity. Say i have a simple table ( even a spreadsheet ) with 3 columns ( sales, cost and time ) and on the rows i have the name of the SKU and we have a 100 rows. I am trying to integrate a sales plan ( which is mostly text ) with lets say the sales projections ( generated by some AI model ) stored in the db.
This is a reasonably appropriate use case in a real world scenario where my knowledge graph would be expected to not only list the salient points of the sales plan BUT also be able to answer queries like "based on the sales projections for AZ for the coming winter, whats the stocking strategy" ( this of course will be answered by an LLM but will connect to the KG for the details and the first part of the query would HAVE to rely on some representation of the numbers within the KG and the 2nd half would use the textual content )
Is there someway KG designers already handle details of numeric data inclusion ? do they just map the DB schema ( for e.g. column names as child entities and "sales plan" as the parent entity with appropriate relation tags ) and then someone needs to query the table to get the answer ? OR can they be connected to stored procedures that are constantly refreshed with pre defined summaries ( like assuming that most users would want to know the region with min projected sales in order to plan marketing campaigns and hence the stored proc would keep this number as a property of the child entity .."title"-> col name ; "min val" ; < other numerical aggregate values > )
sorry but this whole space of structured data and its place in the KG world is a little overwhelming for me , currently