Consider, consider:
UUID,GeoSpatialColumn,name,value,timestamp,type,location,picture,supervisor
1000011365058823906,None,”{None, None, brian, 1.0}”,,”{None, None, 2013-04-04 17:59:58, None}”,”{Type C, None, , 1.0}”,”{Loc B, None, foo, 0.46}; {Loc C, None, foo, 0.46}; {Loc D, None, foo, 0.46}”,”{cugl69808.jpg, None, , 1.0}”,”{None, None, superc, 1.0}”
There are some novel formulations in there that are a function of our data structure. The first is the repeating pattern of:
“{None, None, brian, 1.0}” which represents our “inner” multi-valued knowledge representation. Specifically, it represents the “Vocab Name” (none, it’s not a vocab), the measure (none of these are a measure.), the “freetext”, and the Certainty.
Then there’s the repeating repeating pattern of the “Loc B, Loc C, Loc D” tuples, which represents a user-selected multi-valued attribute.
To a large degree, any given tuple-as-column will likely be filled in consistently by the user, which allows for refinement to happen after the user gets the data.
However, the current csv structure basically supports a boolean “in-text” search for presence/absence of a term, and tabular analysis on that search. For our community, is that level of search acceptable in a CSV? Especially as the “real” data is in the database that can be accessed by basically anything?
If it’s not, what kind of research do you do with a CSV?
Please! Tell me!
