- From: RA Poell <poell@fel.tno.nl>
- Date: Sat, 06 Oct 2001 08:43:55 +0200
- To: Sandro Hawke <sandro@w3.org>, www-rdf-logic@w3.org
- Message-ID: <3BBEA82B.7AA3292@fel.tno.nl>
Sandro Hawke wrote: > > We want to be > able to have semantic web pages be little programs (like HTML pages > with javascript, but cleaner). We want to be able to validate a date > field or do currency conversion, etc, etc. > Just to hook on to these points. The Semantic Web should be machine and human readable. The way I tackle this aspect in Notion System is that starting from the point that the human user is the master of information creation the system must make something from it so it can work with it. For things that are supposed to represent periods of time (like dates, indications of centuries, etc) it uses a service called DatExpert (Sandro: one of the your "little programs") that analyses the input data and transforms it into an agreed standard period exchange format. For DatExpert this is a 64 bit integer with the 0 somewhere 15 billion years ago (a bit beyond the Big Bang) and capable of going 277 billion years in the future. Nothing new with this? Yes there are a few things. 1 the used service has extended knowledge about various calendars and their specific aspects like e.g. the geographical area and time period in which these calendars are used. 2 it might use other information in the semantic network in order to figure out start and end dates from events or statuses that are used as dates (second world war, II Dynasty of Chang). 3 the obtained information is added in two separate properties (derived start "date" and derived end "date") to the original information. For these derived dates again some properties are added like the quality of the conversion, nodes used, service used and a timestamp. This info is for machine exploitation only (but human users can see it if they really want to). Why two dates? Well every date indicates a period over time. Whether the period is to be considered as a point in time only depends on the resolution of the view you have on it (year, second, etc) it doesn't depend on the period itself. This allows you to use upper and lower limits as necessary. 4 When representing dates afterwards to a human user a "displaying service" might either represent the original information and / or use the DatExpert service to reconvert this in the form the particular user is used to see dates. In resume: Original date slots (human created) may contain any kind of value (traditional date forms up to things like "Easter 1960" or "l'an mil trois cent vingt trois ap. J.C."), derived date slots contain machine processable forms (using the French "NumberConversion" service for the last example). Historians, geologists and astronomers can use dates as they are used to also (do you see the resolution of the 4 time scales?) The knowledge (logic) about dates is a combination of facts (e.g. about calendars) within the network and rules how to use these facts. More important even is the fact that if a conversion service is not capable of doing a reliable conversion at time T0 it might be that the user made a mistake (active interaction might correct this) but also that the network lacked some facts. But at time T1 this conversion might succeed (the semantic web is constantly growing). Should we bother the user with this? Personally I would not like that. Accepting original data not confirming the rules is really important. A real life experience: In France an editor published during a few years books with a wrong ISBN (wrong according to the rules applied to ISBN's). The "ISBN" service of Notion System did tell me the ISBN was wrong but as the books are known under these numbers they must be accepted. The use of these kinds of services in the semantic web will be extremely useful in all aspects. Allowing to bridge non acceptable constraints to human users (but helping them to correct possible errors) and still providing functionality as good as possible and capable of exploiting new knowledge as it is added to the sw. For these real life situations I conclude for the semantic web that: - applying strict rules (constraints) at all time might lead to unacceptable situations (ISBN example) - services (like the ones mentioned) must be trusted - derived information (in particular the ones with a low reliability) should be resubmitted periodically in order to improve the quality of the information. - human users remain for the next few years the masters of information creation, machines (applications, services) should be come the masters of transformation for human exploitation. Ronald Poell
Received on Saturday, 6 October 2001 02:44:32 UTC