Creating a standardised, vendor neutral dataset is a rather huge task, I can't see what the commercial incentive would be. I'm also wary of the "creating yet another standard" problem. Arguably, creating KiCad libraries is already a vendor neutral, open source data format.
I have contributed data to the official KiCad libraries, mostly created with scripted generator tools (Python). I wrote a tool called symgen which generates symbols based on a simple text based part description, there are other tools which take a CSV or text file and convert to a symbol, and tools which take a JSON data file and create footprints. There are also parametric generators for 3d models using FreeCad.
The biggest problem with all these tools is just getting the data into any readable form, when the vast majority of data is only published in PDFs. I have tried scraping PDFs but it does not really work. For the more complex devices such as MCU, FPGA, some manufacturers publish text/CSV files which can be parsed easily, or there are tools designed for software configuration which can also provide a text file for symbol creation.
It doesn't matter so much that the data is not a standard form, scripts can do conversion quite easily. So what would be really useful as a first step is to get key parameters from the datasheet to be in a data file, perhaps XML. It seems like the source of that data should be manufacturers.