[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [Elist Home]
Subject: RE: HM.applications-Profiling-Level of Details/Abstraction
Good: a static description. That is representable in an XML Schema given you take out the illegal whitespaces in the generic identifiers (unqualified names). The nut of the problem in the original paper described reasoning by analogy. That is, all the system knows is how to turn the heater on or off. It has to reason its way to facts or really tests to figure out the air conditioner. That is why it has to identify an analog, then do the substitution. It actually doesn't know the air conditioner is there. It knows there is a red button to activate the heater and there is a blue button. One can do testing and one might also compute "certainty/uncertainty" using Bayesian/Demptster-Schafer or belief functions. Seems a bit much for the example but I bring it up because it eventually becomes necessary for other scenarios. Bayesian is usually more effective where conditions and ranges are known (machine systems) and belief systems where there is more uncertainty but expertise exists (medical systems). Going off track: I was after scenarios to help us find some boundaries for HumanML and how it works with other systems. Probably, HumanML is just an organizing domain itself. Len http://www.mp3.com/LenBullard Ekam sat.h, Vipraah bahudhaa vadanti. Daamyata. Datta. Dayadhvam.h -----Original Message----- From: Mark Brownell [mailto:gizmotron@earthlink.net] So for the sake of information like this above, if I understand it correctly, here is a MTML representation of the same information in a human readable form, that could be used by an application using PNLP, Parallel Numerical Lineal Parser. The MTML below is descriptive of a single object, that could be expressive of many objects, that are about things that happen while in a car. <while in the car> -- relates to things that happen while in a car. <human comfort > -- relates to human comfort wherever it may happen. <senses change in comfort> -- relates to senses involving comfort. <reaction to change in comfort> -- implies a human action to change <possible choices in this situation> <air condition controls on the dashboard> <air conditioner on/off switch> <automobile characteristics> <turns heater off when activated/> <is visible on dashboard with a blue indicator/> <is illuminated with a blue indicator light at night/> <turns off air conditioner if selected while AC is on/> </automobile characteristics> </air conditioner on switch> <heater on switch> <automobile characteristics> <turns heater off when activated/> <is visible on dashboard with a blue indicator/> <is illuminated with a blue indicator light at night/> <turns off air conditioner if selected while AC is on/> </automobile characteristics> </heater on switch> </air condition controls on the dashboard> <open windows> <automobile characteristics> <roll down windows using the correct window switch/> </automobile characteristics> </open windows> <do nothing about it> <reference human traits/> <do nothing about it> </possible choices in this situation> </reaction to change in comfort> </senses change in comfort> </human comfort > </while in the car> <while in the car> ... more stuff about cars </while in the car> This set of nested tags could be used to reconstruct sentences that make sense to the inquiring application using this markup. At any level there are objects within other objects. So it is possible to gather together parts of one object that can be used to populate parts of another object. It just depends on how you want to build the application. As you can see the information is in a human readable form.
[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [Elist Home]
Powered by eList eXpress LLC