SemWeb in Real Life - ie not just theory
0 views
Skip to first unread message
Logician
Apr 5, 2008, 10:20:00 AM4/5/08
to Semantic Web
I have read the theory but I cannot find much detail about actually
implementing semantic web theories. Specifically I want to build
relations between bathrooms products. There are about a million
products and the product lines are not text based, ie they are linked
by design which is a visual consideration. They are also linked by
reviews and recommendations, ie a good name.
Using Web 1 type of programming, I have to in effect tag or label
products and then just manually build indexes relating products. What
I plan to do is implement something which describes a product in
enough detail as a data object which is then related automatically.
The actual description is a little hard to do using software, since it
is an image, but I intend to build something which can analyse images,
if not it will be a matter of building a description using user
clicks. ie if a user is searching for cubic taps and they then click
on taps 1,2,5,6,8 buying 5,6 the software will link cubic taps to 5,6.
As the clicks increase for different users a data model will emerge.
Eventually 5,6 will have have a set of relations, eg 5,6 will relate
to taps, cubic taps, high quality taps, good seller, chrome plated,
and also other related products, eg baths, basins. The cross relations
and depth of problem make this an interesting problem. If that is too
easy for anyone, just consider the spare parts and you have ten of
millions of products.
Of course the final solution will be someone will just enter "designer
cubic bathroom" and the software handles the whole query listing
suites (baths, basins, taps) will a final recommendation for a
purchase, or even someone's profile is identified, eg a modern
professional with money and the software makes a recommendation for a
purchase.
This can all be done using basic SQL and an OO language. The problem
is the degree of help from large companies. I do not want something so
bespoke that it is not usable by other people (ie other developers).
So I am really looking for real implementations in terms of languages,
compilers, IDE's, source code, for the theories behind semantic web. I
think if it works well for bathroom goods, then any range of goods not
described well by text, eg clothing, or be indexed.
Does anyone have any salient links for the above requested resources?
implementing semantic web theories. Specifically I want to build
relations between bathrooms products. There are about a million
products and the product lines are not text based, ie they are linked
by design which is a visual consideration. They are also linked by
reviews and recommendations, ie a good name.
Using Web 1 type of programming, I have to in effect tag or label
products and then just manually build indexes relating products. What
I plan to do is implement something which describes a product in
enough detail as a data object which is then related automatically.
The actual description is a little hard to do using software, since it
is an image, but I intend to build something which can analyse images,
if not it will be a matter of building a description using user
clicks. ie if a user is searching for cubic taps and they then click
on taps 1,2,5,6,8 buying 5,6 the software will link cubic taps to 5,6.
As the clicks increase for different users a data model will emerge.
Eventually 5,6 will have have a set of relations, eg 5,6 will relate
to taps, cubic taps, high quality taps, good seller, chrome plated,
and also other related products, eg baths, basins. The cross relations
and depth of problem make this an interesting problem. If that is too
easy for anyone, just consider the spare parts and you have ten of
millions of products.
Of course the final solution will be someone will just enter "designer
cubic bathroom" and the software handles the whole query listing
suites (baths, basins, taps) will a final recommendation for a
purchase, or even someone's profile is identified, eg a modern
professional with money and the software makes a recommendation for a
purchase.
This can all be done using basic SQL and an OO language. The problem
is the degree of help from large companies. I do not want something so
bespoke that it is not usable by other people (ie other developers).
So I am really looking for real implementations in terms of languages,
compilers, IDE's, source code, for the theories behind semantic web. I
think if it works well for bathroom goods, then any range of goods not
described well by text, eg clothing, or be indexed.
Does anyone have any salient links for the above requested resources?
Adrian Walker
Apr 5, 2008, 12:02:34 PM4/5/08
to semant...@googlegroups.com
Hi Logician --
You wrote
This [task] can all be done using basic SQL and an OO language. The problem is the degree of help from large companies. I do not want something so
bespoke that it is not usable by other people.
You may be interested in the the system online at the site below for this purpose.
It's a kind of Wiki, for writing and running applications as business rules. The rules are in open vocabulary, largely open syntax, English. Shared use is free.
From the business rules, the system automatically generates and runs SQL. In many cases, the SQL is too complex to be written reliably by hand. However, the system explains the results, in English, at the business level. The explanations can also be viewed as plans.
There's a simple browser-based author- and user-interface. The system can also be used as an SOA endpoint, by downloading and extending a provided Java stub.
Some background is in [1,2,3].
I hope this may be of interest. We can discuss off-list if you wish.
-- Adrian
Internet Business Logic
A Wiki and SOA Endpoint for Executable Open Vocabulary English over SQL
Online at www.reengineeringllc.com Shared use is free
Adrian Walker
Reengineering
[1] www.reengineeringllc.com/A_Wiki_for_Business_Rules_in_Open_Vocabulary_Executable_English.pdf
[2] www.reengineeringllc.com/Oil_Industry_Supply_Chain_by_Kowalski_and_Walker.pdf
[3] www.reengineeringllc.com/ibldrugdbdemo1.htm (Flash video with audio)
You wrote
This [task] can all be done using basic SQL and an OO language. The problem is the degree of help from large companies. I do not want something so
bespoke that it is not usable by other people.
You may be interested in the the system online at the site below for this purpose.
It's a kind of Wiki, for writing and running applications as business rules. The rules are in open vocabulary, largely open syntax, English. Shared use is free.
From the business rules, the system automatically generates and runs SQL. In many cases, the SQL is too complex to be written reliably by hand. However, the system explains the results, in English, at the business level. The explanations can also be viewed as plans.
There's a simple browser-based author- and user-interface. The system can also be used as an SOA endpoint, by downloading and extending a provided Java stub.
Some background is in [1,2,3].
I hope this may be of interest. We can discuss off-list if you wish.
-- Adrian
Internet Business Logic
A Wiki and SOA Endpoint for Executable Open Vocabulary English over SQL
Online at www.reengineeringllc.com Shared use is free
Adrian Walker
Reengineering
[1] www.reengineeringllc.com/A_Wiki_for_Business_Rules_in_Open_Vocabulary_Executable_English.pdf
[2] www.reengineeringllc.com/Oil_Industry_Supply_Chain_by_Kowalski_and_Walker.pdf
[3] www.reengineeringllc.com/ibldrugdbdemo1.htm (Flash video with audio)
Logician
Apr 5, 2008, 1:36:57 PM4/5/08
to Semantic Web
On Apr 5, 5:02 pm, "Adrian Walker" <adriandwal...@gmail.com> wrote:
> Hi Logician --
>
> You wrote
>
> is the degree of help from large companies. I do not want something so
> bespoke that it is not usable by other people.*
>
> You may be interested in the the system online at the site below for this
> purpose.
>
> It's a kind of Wiki, for writing and running applications as business
> rules. The rules are in open vocabulary, largely open syntax, English.
> Shared use is free.
>
> From the business rules, the system automatically generates and runs SQL.
> In many cases, the SQL is too complex to be written reliably by hand.
> However, the system explains the results, in English, at the business
> level. The explanations can also be viewed as plans.
>
> There's a simple browser-based author- and user-interface. The system can
> also be used as an SOA endpoint, by downloading and extending a provided
> Java stub.
>
> Some background is in [1,2,3].
>
> I hope this may be of interest. We can discuss off-list if you wish.
I read your link. I can see it may have some uses, but the issues I
> You may be interested in the the system online at the site below for this
> purpose.
>
> It's a kind of Wiki, for writing and running applications as business
> rules. The rules are in open vocabulary, largely open syntax, English.
> Shared use is free.
>
> From the business rules, the system automatically generates and runs SQL.
> In many cases, the SQL is too complex to be written reliably by hand.
> However, the system explains the results, in English, at the business
> level. The explanations can also be viewed as plans.
>
> There's a simple browser-based author- and user-interface. The system can
> also be used as an SOA endpoint, by downloading and extending a provided
> Java stub.
>
> Some background is in [1,2,3].
>
> I hope this may be of interest. We can discuss off-list if you wish.
have are more related to actually getting the data which is basically
from user actions and then the data is related.
Your link seems to generate queries but the assumption is made that
the data already exists. It is actually getting the data which is very
hard.
The issue of language, or semantics, in my business has a different
meaning to just understanding words. The difference is that products
are described by design criteria, quality, etc. It is not a matter of
understanding English words and translating into a machine language
(eg SQL) but of actually understanding the terms and the data related
to them, almost like a new language.
The key difference is that in sales qualitative issues matter. There
are few facts about a product. If I say High Quality it is not a fact
but a view. If several people have the view, it is an accepted view. A
parallel would be to invent a language to describe something because
words do not exist. Inventing the language would be hard but once
invented it could be used to describe all similar objects.
I agree once that is done, you could use your kind of thinking to them
find different ways of saying High Quality, eg good products,
recommended, etc
To clarify: Microsoft invested in supported XML with classes in C#
making it easy to use XML. Now we have RDF and recommendations to
extend XML, I assume Microsoft will invest in more C# classes so we
can easily process RDF.
I think once key difference is that semantics related to sales is a
business, and not actually overhead (eg accounts is overhead). The key
to distribution is product knowledge. If such knowledge can be changed
then the entire retail systems of the world can change allowing people
to find what they want almost instantly. This has an impact of
billions as billions are used in advertising, brochures, shop
displays, sales calls, branding, etc.
>
> -- Adrian
>
> Internet Business Logic
> A Wiki and SOA Endpoint for Executable Open Vocabulary English over SQL
>
> Adrian Walker
> Reengineering
>
> [1]www.reengineeringllc.com/A_Wiki_for_Business_Rules_in_Open_Vocabulary...
>
> [2]www.reengineeringllc.com/Oil_Industry_Supply_Chain_by_Kowalski_and_Wa...
>
> - Show quoted text -
knorth
Apr 8, 2008, 9:06:16 PM4/8/08
to Semantic Web
> This can all be done using basic SQL and an OO language.
..
> So I am really looking for real implementations in terms of languages,
> compilers, IDE's, source code, for the theories behind semantic web. I
> think if it works well for bathroom goods, then any range of goods not
> described well by text, eg clothing, or be indexed.
>
> Does anyone have any salient links for the above requested resources?
Sometimes we don't find information because of differences in
> compilers, IDE's, source code, for the theories behind semantic web. I
> think if it works well for bathroom goods, then any range of goods not
> described well by text, eg clothing, or be indexed.
>
> Does anyone have any salient links for the above requested resources?
terminology.
For some "real world" implementations, look at sites such as Twine
(Radar Networks) and OpenCalais (Reuters):
http://www.twine.com/
http://www.opencalais.com/
The move to the next generation web has been described as moving from
a web of linked documents to a web of linked data - so the operative
phrase you're looking for is "linked data" - such as integrating data
from SQL databases for publication on the web.
One of the instructors for the workshop at LinkedData Planet is from
Oracle. She'll explain a system for customer subscription information
for a mobile multimedia service provider. The conference runs June
17-18 in New York City:
http://www.linkeddataplanet.com
Danny Ayers
Apr 9, 2008, 5:05:32 AM4/9/08
to semant...@googlegroups.com, SWIG
[cc'ing semant...@w3.org, original thread at http://groups.google.com/group/semantic_web/browse_thread/thread/299e6336e05f5188 ]
That actually sounds the kind of job for which RDF and OWL (the Web Ontology Language) are well suited, though offhand I can't think of any existing vocabularies in that space.
Right, the nearest thing I know of to that is the Review vocabulary:
Microsoft haven't exactly been quick off the mark for RDF support (though their ADO.Net Entity Framework stuff is very RDF-like, but unfortunately doesn't use URIs as identifiers). Fortunately there are loads of open source toolkits available, see:
http://www4.wiwiss.fu-berlin.de/bizer/toolkits/
Absolutely!
Not long ago I started looking into product description for a considerably simpler scenario than you describe (for music equipment, the main complication was representing part-whole relationships for guitar parts etc). I was surprised to find how little work had apparently been done in this area, given the commercial potential.
For what it's worth I got as far as thinking something very like the core terms of FRBR would make a good starting point, borrowing the part-whole stuff from somewhere like Cyc or maybe one of the proposed Upper Ontologies, the whole lot being liberally sprinkled with SKOW.
Cheers,
Danny.
--
http://dannyayers.com
~
http://blogs.talis.com/nodalities/this_weeks_semantic_web/
The issue of language, or semantics, in my business has a different
meaning to just understanding words. The difference is that products
are described by design criteria, quality, etc. It is not a matter of
understanding English words and translating into a machine language
(eg SQL) but of actually understanding the terms and the data related
to them, almost like a new language.
That actually sounds the kind of job for which RDF and OWL (the Web Ontology Language) are well suited, though offhand I can't think of any existing vocabularies in that space.
The key difference is that in sales qualitative issues matter. There
are few facts about a product. If I say High Quality it is not a fact
but a view. If several people have the view, it is an accepted view. A
parallel would be to invent a language to describe something because
words do not exist. Inventing the language would be hard but once
invented it could be used to describe all similar objects.
Right, the nearest thing I know of to that is the Review vocabulary:
http://purl.org/stuff/rev#
which is used under the hood at http://revyu.com
I agree once that is done, you could use your kind of thinking to them
find different ways of saying High Quality, eg good products,
recommended, etc
To clarify: Microsoft invested in supported XML with classes in C#
making it easy to use XML. Now we have RDF and recommendations to
extend XML, I assume Microsoft will invest in more C# classes so we
can easily process RDF.
Microsoft haven't exactly been quick off the mark for RDF support (though their ADO.Net Entity Framework stuff is very RDF-like, but unfortunately doesn't use URIs as identifiers). Fortunately there are loads of open source toolkits available, see:
http://www4.wiwiss.fu-berlin.de/bizer/toolkits/
I think once key difference is that semantics related to sales is a
business, and not actually overhead (eg accounts is overhead). The key
to distribution is product knowledge. If such knowledge can be changed
then the entire retail systems of the world can change allowing people
to find what they want almost instantly. This has an impact of
billions as billions are used in advertising, brochures, shop
displays, sales calls, branding, etc.
Absolutely!
Not long ago I started looking into product description for a considerably simpler scenario than you describe (for music equipment, the main complication was representing part-whole relationships for guitar parts etc). I was surprised to find how little work had apparently been done in this area, given the commercial potential.
For what it's worth I got as far as thinking something very like the core terms of FRBR would make a good starting point, borrowing the part-whole stuff from somewhere like Cyc or maybe one of the proposed Upper Ontologies, the whole lot being liberally sprinkled with SKOW.
Danny.
--
http://dannyayers.com
~
http://blogs.talis.com/nodalities/this_weeks_semantic_web/
Logician
Apr 11, 2008, 12:58:52 AM4/11/08
to Semantic Web
On Apr 9, 10:05 am, "Danny Ayers" <danny.ay...@gmail.com> wrote:
> [cc'ing semantic-...@w3.org, original thread athttp://groups.google.com/group/semantic_web/browse_thread/thread/299e...]
>
> On 05/04/2008, Logician <sa...@logicians.com> wrote:
>
>
>
> > The issue of language, or semantics, in my business has a different
> > meaning to just understanding words. The difference is that products
> > are described by design criteria, quality, etc. It is not a matter of
> > understanding English words and translating into a machine language
> > (eg SQL) but of actually understanding the terms and the data related
> > to them, almost like a new language.
>
> That actually sounds the kind of job for which RDF and OWL (the Web
> Ontology Language) are well suited, though offhand I can't think of any
> existing vocabularies in that space.
>
> The key difference is that in sales qualitative issues matter. There
>
> > are few facts about a product. If I say High Quality it is not a fact
> > but a view. If several people have the view, it is an accepted view. A
> > parallel would be to invent a language to describe something because
> > words do not exist. Inventing the language would be hard but once
> > invented it could be used to describe all similar objects.
>
> Right, the nearest thing I know of to that is the Review vocabulary:
>
> http://purl.org/stuff/rev#
>
> which is used under the hood athttp://revyu.com
> On 05/04/2008, Logician <sa...@logicians.com> wrote:
>
>
>
> > The issue of language, or semantics, in my business has a different
> > meaning to just understanding words. The difference is that products
> > are described by design criteria, quality, etc. It is not a matter of
> > understanding English words and translating into a machine language
> > (eg SQL) but of actually understanding the terms and the data related
> > to them, almost like a new language.
>
> That actually sounds the kind of job for which RDF and OWL (the Web
> Ontology Language) are well suited, though offhand I can't think of any
> existing vocabularies in that space.
>
> The key difference is that in sales qualitative issues matter. There
>
> > are few facts about a product. If I say High Quality it is not a fact
> > but a view. If several people have the view, it is an accepted view. A
> > parallel would be to invent a language to describe something because
> > words do not exist. Inventing the language would be hard but once
> > invented it could be used to describe all similar objects.
>
> Right, the nearest thing I know of to that is the Review vocabulary:
>
> http://purl.org/stuff/rev#
>
>
> I agree once that is done, you could use your kind of thinking to them
>
> > find different ways of saying High Quality, eg good products,
> > recommended, etc
>
> > To clarify: Microsoft invested in supported XML with classes in C#
> > making it easy to use XML. Now we have RDF and recommendations to
> > extend XML, I assume Microsoft will invest in more C# classes so we
> > can easily process RDF.
>
> Microsoft haven't exactly been quick off the mark for RDF support (though
> their ADO.Net Entity Framework stuff is very RDF-like, but unfortunately
> doesn't use URIs as identifiers).
Your are right about MS and RDF. The pity is that MS is great at
> I agree once that is done, you could use your kind of thinking to them
>
> > find different ways of saying High Quality, eg good products,
> > recommended, etc
>
> > To clarify: Microsoft invested in supported XML with classes in C#
> > making it easy to use XML. Now we have RDF and recommendations to
> > extend XML, I assume Microsoft will invest in more C# classes so we
> > can easily process RDF.
>
> Microsoft haven't exactly been quick off the mark for RDF support (though
> their ADO.Net Entity Framework stuff is very RDF-like, but unfortunately
> doesn't use URIs as identifiers).
writing an IDE. Anyone who has used VC# knows that MS has a great
product.
Sadly, MS is very slow to adopt outside technologies. It has some good
classes for XML in C# but the company has not done the integration
that easily (but it works).
What makes my situation so different is that I will directly be using
data to make money via a form of e-commerce.
I go to meetup sessions (meetup.com) sometimes and I am often asked
what is different to my ideas to Pricerunner (which purports to
represent the best deals using links). The difference is that
Pricerunner and similar sites do not show relations between products
based on sales criteria. The same drawback applies to almost all
websites engaged in selling (direct or indirect sales).
It is very easy to list products. It is also not hard to relate
products based on user traffic but we as developers need the product
software. So where is MS on the issue?
Normally data is used internally as is seen as an overhead by
business executives. This always has meant that new technologies are
only slowed adopted by business executives.
The real issue is what will happen to search engines when web data
controls wen traffic and not web links? SInce google is based on a
link popularity method, what will happen when data itself and user
traffic starts to control the web?
I think Web 3 will really change the web a lot. Will Google survive in
a space that has seen so many go (don't buy too many Google shares)?
Recall when everyone used Yahoo! for search - that changed very
quickly? No one in 1996 would have said in 2006 Yahoo! would be
failing in search!
Reply all
Reply to author
Forward
0 new messages