Microsoft recently announced  this new "data type", which is actually a connection between input in Excel and the proprietary "knowledge graph", and users start to ask for equivalents  (the MS posting was prior to 1st of April).
The user stories are like this
* Eve enters "Country" in cell A1 (and some countries below like USA, Canada, Spain, Japan...) and "Population" in cell B1 expecting the function to retrieve the values from an external source.
* Eve takes names and addresses of her customers and wants to retrieve the current area codes from an external source.
* Eve enters "Year" in A1 (and 2018..2010 below) and "Football Champion" in B1 and expects the winning teams of her country.
* Eve enters "Element" in A1 and expects a dropdown list of external information in B1 where she can choose "Molecular Mass".
* Eve wants to know where exactly the external data come from and on what basis it was collected.
(Dynamical data like stock value or exchange rates are a different type of information.)
In terms of open source such a feature would be a query to Wikidata (the sidebar facts at Wikipedia), or any other open and transparent data provider, which is technically pretty simple. The challenge is to get the right data out of the external source accepting fuzzy input (like USA instead of United States of America, or to understand Element as Chemical Element) and the interesting information itself (the prioritization of locale information in the second example is artificially).
I think such a function goes beyond the actual scope and should be implemented as an extension. But that's up for discussion as such data retrieval with full transparency would be a killer feature.
PS: The feature is compared with Lotus Improv  and Jvelin  but I cannot follow this derivation idea.
Javelin and Improv were financial forecasting/modeling tools. Their secondary function was general business intelligence.
They required data input to adhere to a very specific set of rules.
Basically, it was what, when, where, and numbers. The specific activity was irrelevant. Consequently, it could be used with equal ease to predict the number of deaths in a specific location each month, or the future gross profit for a specific location in a specific month.
If your data was periodic, and you needed periodic based forecasts, the tool was, for you, mandatory.
If either your data was not periodic, or you did not need periodic based forecasts, the tool was, for you, useless.
The simplest route for LibO to clone Javelin, would be to rewrite the extension that enables R functionality within Calc, such that it automatically installs all of the needed dependencies, then write a _LibO with R_ manual, using examples similar to those in the original Javelin documentation, as well as more recent Business Intelligence examples.
(In reply to jonathon from comment #1)
> The simplest route for LibO to clone Javelin...
While R surely boosts the functionality of any application that's off-topic here. The request is the integration of Wikidata.
Interesting proposal. Needs breaking down into implementable bits and pieces, preferably via extensions, preferably using existing open data / open knowledge libraries.
Could seed OKFN / Jugend hackt / hackfest event ideas.
No further input from UX needed by now. Maybe when it comes to the implementation.