- From: Bill Roberts <bill@swirrl.com>
- Date: Fri, 5 Apr 2013 10:22:09 +0100
- To: "public-gld-comments@w3.org" <public-gld-comments@w3.org>
- Message-Id: <16E4E445-4FE3-4389-87BF-59195E562C42@swirrl.com>
I've also found the DCAT document very useful in providing advice on useful standardised metadata terms for describing linked data datasets. (I realise that DCAT is applicable to all kinds of datasets in data catalogues, but I have been applying it in a linked data context). I tend to use a combination of DCAT and VOID to describe such datasets - and generally always provide the minimum set of metadata described in the Data Cube draft (WD-vocab-data-cube-20130312 section 9) plus a few other useful items. My comments on the Data Cube document mentioned the usefulness of coverage and granularity information (spatial and temporal) for describing datasets appearing in a catalogue. This is particularly relevant for data cube datasets, but not only datacube datasets. dct:spatial and dct:temporal handle the coverage question - that's fine. Granularity is also a useful concept though - if you are looking for say crime statistics, there's a big difference in number of crimes for the whole country versus number of crimes per local government region or crimes per street. This may be a complicated concept to describe reliably - in the data cube case, an option might be to use (as the value of some kind of granularity predicate) the class of the smallest type of area used in the values of the area dimension. In other kinds of dataset it may be harder to define precisely. Similarly, temporal granularity is of interest - whether data is provided annually, monthly, daily etc. This is a separate concept from accrualPeriodicity as it's about the contents of the data not the frequency of update of the dataset. Do the authors have ideas for potential standard terms to describe spatial and temporal granularity? One other question: the range of accrualPeriodicity is a Frequency. Looking into this recently, I found it hard to find a good set of standard URIs for frequencies (yearly, quarterly etc). Can the authors recommend a standard set for use in this context? Would it be easier if the predicate was rephrased as a period (updateInterval or something) rather than a frequency, as there are common terms for year, month etc that could be used as values. I would also echo other recent comments on the list that a way of describing an API associated with a dataset would be useful, as a web service or API has some different characteristics to a downloadable file and to a certain extent needs different ways of describing it. Best regards Bill Bill Roberts, CEO, Swirrl IT Limited Mobile: +44 (0) 7717 160378 | Skype: billroberts1966 | @billroberts http://swirrl.com | http://publishmydata.com
Received on Friday, 5 April 2013 09:22:46 UTC