W3C home > Mailing lists > Public > semantic-web@w3.org > September 2010

Re: ANN: The Vehicle Sales Ontology - Cars, Bikes, Boats on the Web of Data - http://purl.org/vso/ns

From: Michael F Uschold <uschold@gmail.com>
Date: Thu, 9 Sep 2010 12:16:10 -0700
Message-ID: <AANLkTikRDPCwwnXQBT-2z1CKvwEJCbvHBxHEaka82fT0@mail.gmail.com>
To: Martin Hepp <martin.hepp@ebusiness-unibw.org>
Cc: semantic-web@w3.org, Pat Hayes <phayes@ihmc.us>
Martin, you make good points.

In summary,

   1. VSO is not messy
   2. VSO is only far from common sense insofar as not assuming that a kayak
   is a boat is far from common sense.
   3. the main use of the ontology being search should be a critical factor
   in making ontology design choices hence:
      1. it is not useful to model all kinds of manual power (paddle,oar
      etc) explicitly, but it is important to have these kind of craft in the
      ontology if they are commonly rented.
      2. not assuming a kayak is a boat may cause minimal harm
      3. not assuming a kayak is a boat should be linked to a benefit to
      users?
      What is that benefit?

More comments below.

Michael



On Thu, Sep 9, 2010 at 9:17 AM, Martin Hepp <martin.hepp@ebusiness-unibw.org
> wrote:

> Hi Michael,
> I agree with your trade-offs. However, just clarifying my point: The VSO
> does simply not say that a Kayak need to be considered a "boat". It does not
> make a kayak a non-boat. Such is IMO also following the principle of minimal
> ontological commitment.


This is a good principle, but I think it does not apply here. There is no
real world argument (that I am aware of) that says a kayak might not be a
boat, (nor for example that a brick might not be a physical object). The
main argument seems to be one of convenience for domains and ranges.

>
>
>  IMHO it is dangerous to stray far from common sense - it is too expensive
>>> in scaring users away by getting surprising and incorrect behavior in
>>> applications.
>>>
>> I assume you are not stating that the VSO is "far from common sense", are
> you?


This comment is not about VSO overall.   VSO is only far from common sense
insofar as not assuming that a kayak is a boat is far from common sense.   I
originally thought you were saying saying a kayak is *not *a boat -- which
is further from common sense than what VSO says.


>
>
>  IMHO catering too much to the whims of inference engines and making life
>>> easy for web developers often results in an unfortunate amount of messiness
>>> in an ontology.
>>>
>> I assume you are not stating that the VSO is "messy", are you?
>


No, this is a general remark about workarounds (e.g. to stay within OWL-DL)
which are rarely pretty -- even when the workarounds are very thoughtfully
designed and the need is clearly justified.

In the case of VSO, placing inferencing and domain/range convenience as a
high priority seemed to result in the [conceptual] workaround: "do not
assume that a kayak is a boat".  This is an omission, not  messy.

I know that having domains and ranges can be very useful for UI (by limiting
choices to only the sensible ones) -- so this may be an important design
decision.  However, it struck me that the problem with domains and ranges by
putting kayak under boat arose from [wrongly] assuming that a boat was
motorizable.


>
> Side note: The relevance of a strict class hierarchy for Web ontologies is,
> IMO, very much overrated in Semantic Web research. The top-level ontological
> distinctions are very important for reusing data (e.g. keeping book titles
> and book copies apart), but not for search.
> Who will search for things as generic as a "human-powered arm vehicle"? If
> you want to rent a kayak, you will search for kayaks and not any type of
> boat.
>
> So I would say that the subject of discussion - whether kayaks are always
> boats or not - it not a particularly important axiom.
>
>
This is an extremely important point: the ontology is mainly going to be
used for search in real business use, so VSO should be optimized for this
purpose.

Kayak rental is a significant business, in some places (e.g. Seattle,
Alaska, maybe Costa Rica and other tourist destinations). In the US it
generally costs more to rent a sea kayak than a car.  So it is important to
have this in your ontology. Paddle surfboards are also rented.

I agree that most people looking for a kayak will type "kayak".  However,
there is a risk that a user may browse under boat, not see kayak and not be
thinking about kayaks, hence not  search for it. Lost revenue.

I think we could all agree that IF there was no cost, then we would just put
kayak under boat and call it good.  The cost seems to be making domains and
ranges awkward for inference engines.

How is this cost linked to benefit to users?

Michael


Martin
>
>
> On 09.09.2010, at 06:09, Pat Hayes wrote:
>
>
>> On Sep 8, 2010, at 4:26 PM, Michael F Uschold wrote:
>>
>>  There are various tradeoffs:
>>>         Conceptual simplicity of the ontology,
>>>                 for easy understandabilty
>>>                 for ease of use
>>>         Alignment with common sense
>>>                 for easy understandabilty
>>>                 to avoid repelling potential users
>>>         Ontological correctness which should correlate with 2. but may
>>> be at odds with 1.
>>>                 to align with common sense
>>>                 to increase correctness and scope of usability
>>>                 more correct can often mean more complex
>>>         Keeping things nice for inference engines
>>>                 to improve functionality in an application
>>>         Keeping things nice for [semantic] web developers and
>>> programmers
>>>                 to encourage use
>>> IMHO it is dangerous to stray far from common sense - it is too expensive
>>> in scaring users away by getting surprising and incorrect behavior in
>>> applications.
>>>
>>> IMHO catering too much to the whims of inference engines and making life
>>> easy for web developers often results in an unfortunate amount of messiness
>>> in an ontology.
>>>
>>> On Mon, Aug 30, 2010 at 11:36 PM, Martin Hepp <
>>> martin.hepp@ebusiness-unibw.org> wrote:
>>> Hi Michael,
>>> Thanks for the feedback! I chose to consider Kayaks as a special kind of
>>> watercraft in order to exclude them from the domain of all properties
>>> associated with having a motor, because to my knowledge, kayaks are
>>> extremely rarely powered by an engine (other than e.g. canoes, which
>>> occasionally have small electric or combustion engines).
>>>
>>> This seems to be trading off ontologically correctness and alignment with
>>> common sense for simplicity and making things nice for web programmers and
>>> inference engines. I am all in favor of simplicity to encourage use, but I
>>> think it is dangerous to be too out of alignment with common sense - which
>>> says a kayak is a boat.
>>>
>>>
>>> By simply also making kayak a subclass of boat, one would recommend (*) a
>>> lot of properties that 99.9% of the kayaks in the world don't have (engine
>>> displacement etc.), which will irritate potential adopters.
>>>
>>> Otherwise, I would have needed rather "ontology expert" classes like
>>> "motorizableWaterVehicle" etc. This would also require complex class
>>> definitions for the range / domain definitions, which cause practical
>>> problems in many pure RDF and RDFS environments (e.g. resolving unionOf
>>> without an OWL reasoner is a pain for developers).
>>>
>>> You seem to be using the term "boat" to mean "motorizable water vehicle".
>>>   I agree that "motorizable water vehicle" is not very natural or simple,
>>> even if it is ontologically correct.
>>>
>>
>> Um... don't the (natural) categories of powered boat, sailboat and rowboat
>> capture the needed distinctions here quite adequately? I have no idea
>> whether a kayak is usually called a rowing boat, but it clearly belongs in a
>> natural class of boats propelled by arm muscles holding a paddle or oar.
>> BTW, there are, or once were, sea-going ships in all these three categories.
>>
>> Pat Hayes
>>
>>
>>  I agree that there can be domain and range issues.  I always cringe at
>>> examples like this where we cater to the whims of the language and the
>>> reasonsers at the cost of common sense.
>>>
>>> You have thought about it longer than I, but intuition and experience
>>> suggests there is likely a way to keep things reasonably simple and also
>>> aligned with common sense.
>>>
>>>
>>>
>>> So it was really just a decision for bringing order to the domains and
>>> ranges of typical properties.
>>>
>>> I hope this modeling compromise is acceptable for all kayakers in the
>>> world.
>>>
>>> Martin
>>>
>>> (*) I am well aware of the specific semantics of rdfs:range and
>>> rdfs:domain ;-)
>>> (**) I know that you know that I am not saying that a kayak is not a boat
>>> but just that a kayak does not always need to be a regular boat for
>>> everybody ;-)
>>>
>>>
>>> On 30.08.2010, at 20:36, Michael F Uschold wrote:
>>>
>>> Overall this ontology is just fine, highly suitable for its intended
>>> purpose.  I do have one [hopefully] minor concern. Why is a kayak not a kind
>>> of a boat? The classification in this ontology goes like this:
>>>        Watercraft
>>>                Boat
>>>                Kayak
>>>                Ship
>>> The source of this (IHMO) mistake may be in the WIkipedia entry for
>>> Watercraft:
>>>
>>> However, there are a number of craft which many people would consider
>>> neither a ship nor a boat, such as:canoes, kayaks, rafts, barges,
>>> catamarans, hydrofoils, windsurfers, surfboards (when used as a paddle
>>> board), jet skis, underwater robots, seaplanes, and torpedoes.
>>>
>>> Contradictorily, the opening words in the definition of kayak in
>>> Wikipedia clearly state that a kayak is a boat:
>>> A kayak (sometimes generalised as a canoe) is a small human-powered boat
>>> that traditionally has a covered deck, and one or more cockpits, each
>>> seating one paddler who strokes a double-bladed paddle.
>>>
>>> I have been a kayaker for 35 years, and every kayaker I know thinks and
>>> speaks of their kayak as a kind of boat. In the US, most whitewater kayakers
>>> consider themselves boaters.
>>>
>>> What competency question justifies this classification?
>>> What is an example of a kayak that is categorically not a boat?
>>>
>>> Michael
>>>
>>>
>>>
>>> On Mon, Aug 30, 2010 at 10:22 AM, Martin Hepp <
>>> martin.hepp@ebusiness-unibw.org> wrote:
>>> Dear all:
>>>
>>> I am happy to announce the first mature release of the Vehicle Sales
>>> Ontology [1], a GoodRelations-compliant [2,3] Web vocabulary for
>>> - Cars,
>>> - Bikes,
>>> - Boats,
>>> - etc.
>>> on the Web of Data.
>>>
>>> It can be used by car listing sites, bike or canoe rental services and
>>> the like.
>>>
>>> In combination with
>>>
>>> - http://purl.org/goodrelations/v1#owns and
>>> - http://purl.org/goodrelations/v1#seeks ,
>>>
>>> it is also possible to expose ownership ("I own a Volkswagen Golf") as
>>> part of online identity data or purchasing interest ("I am looking for a
>>> canoe").
>>>
>>> The ontology recommends DBPedia resource URIs as predefined qualitative
>>> values as much as possible.
>>>
>>> Any feedback is very welcome.
>>>
>>> Best wishes
>>>
>>> Martin
>>> [1] http://purl.org/vso/ns
>>> [2] http://purl.org/goodrelations/v1
>>> [3] http://www.ebusiness-unibw.org/wiki/Own_GoodRelations_Vocabularies
>>>
>>>
>>> --------------------------------------------------------
>>> martin hepp
>>> e-mail:  mhepp@computer.org
>>> www:     http://www.heppnetz.de/
>>> skype:   mfhepp
>>> twitter: mfhepp
>>>
>>> Check out GoodRelations for E-Commerce on the Web of Linked Data!
>>> =================================================================
>>> * Project Main Page: http://purl.org/goodrelations/
>>> * Quickstart Guide for Developers: http://bit.ly/quickstart4gr
>>> * Vocabulary Reference: http://purl.org/goodrelations/v1
>>> * Developer's Wiki: http://www.ebusiness-unibw.org/wiki/GoodRelations
>>> * Examples: http://bit.ly/cookbook4gr
>>> * Presentations: http://bit.ly/grtalks
>>> * Videos: http://bit.ly/grvideos
>>>
>>>
>>>
>>>
>>> --
>>> Michael Uschold, PhD
>>>  LinkedIn: http://tr.im/limfu
>>>  Skype: UscholdM
>>>
>>>
>>>
>>>
>>> --
>>> Michael Uschold, PhD
>>>   LinkedIn: http://tr.im/limfu
>>>   Skype: UscholdM
>>>
>>
>> ------------------------------------------------------------
>> IHMC                                     (850)434 8903 or (650)494 3973
>> 40 South Alcaniz St.           (850)202 4416   office
>> Pensacola                            (850)202 4440   fax
>> FL 32502                              (850)291 0667   mobile
>> phayesAT-SIGNihmc.us       http://www.ihmc.us/users/phayes
>>
>>
>>
>>
>>
>


-- 
Michael Uschold, PhD
   LinkedIn: http://tr.im/limfu
   Skype: UscholdM
Received on Thursday, 9 September 2010 19:16:44 GMT

This archive was generated by hypermail 2.3.1 : Tuesday, 26 March 2013 21:45:38 GMT