W3C home > Mailing lists > Public > public-semweb-lifesci@w3.org > May 2008

Re: Cross Language with topic maps [was Re: 答复: KB note]

From: Jack Park <jack.park@sri.com>
Date: Wed, 28 May 2008 14:32:42 -0700
Message-ID: <483DCF7A.1070402@sri.com>
To: w3c semweb hcls <public-semweb-lifesci@w3.org>
CC: Tim Clark <twclark@nmr.mgh.harvard.edu>

Thanks for the question, Kai.

Topic maps in RDF:
There exists an OWL DL variant of the XML topic maps standard (XTM) [1]
There may be others. Certainly a search on "owl topic map" reveals interest.

I have an OWL Full variant of the TMRM "subject maps" standard, ISO 
13250-5. The TMRM (topic maps reference model) departs from the XML 
topic maps standard (which I helped create) by reducing topic mapping to 
its core functionality, releasing the standard from XML serialization 
schemes. We are now back to the frame-based representations many people 
are familiar with. But, I have yet to find a way to use DL because the 
properties of each OWL class or instance (we call them "subjects") must 
also be subjects; the co-reference of a property as a class instance has 
evaded me: I have yet to make it validate. But, validation as OWL Full 
flys quite easily. Each property instance is then contained in a p-list 
rather than thinking in terms of restrictions. Wierd, but it does answer 
SPARQL queries.

Topic maps do not and should not interfere with the semantic web; they 
bring other characteristics to knowledge representation and organization 
that, I think, enhance the semantic web's capabilities. For one thing, 
they offer two opportunities: they automatically organize all entities 
of the semantic web in a subject-centric fashion, and, in so doing, they 
offer the opportunity to federate (a kind of merge) ontologies of all 
kinds. For instance, it may be that the NCI cancer ontology has a few 
terms (classes) that are also found (perhaps even under a different 
name) in, say, a Parkinson's ontology. We can merge all ontologies, no 
matter the names, so long as we are able to identify that different 
classes happen to be representing the same subject (entity, concept, 
whatever you wish to call it), the two entities are merged. Those that 
don't merge remain still linked in their graphs. This allows for 
potential "aha!" moments (some might say "Black Swan Events"), as an 
emergent behavior. I don't mean to sound like I'm making arm-waving 
claims; I am merely sketching some possibilities. It's also the case 
that you don't need topic maps to do that. Topic maps represent, more 
than anything, a slightly different way to think about the problem space.

[1] http://xml.coverpages.org/CreganTMs-OWL200505.pdf

Kei Cheung wrote:
> Hi Eric et al,
> I'm glad that umls, topic map, ... were mentioned. We have to do more 
> than literal translation or linguistics. It's semantics!
> Traditional Chinese medicine embodies rich dialectical thought, such as 
> that of the holistic connections and the unity of yin and yang. It deals 
> with many facets of human anatomy and physiology: 臟腑 zang-fu (organs), 
> 穴 meridians (main and collateral channels), 氣 qi (vital energy), 血 
> blood, *靜 *jing (essence of life), body fluid, the inside and outside 
> of the body, as well as the connections between the whole and the parts.
> I wonder if there is a Chinese counterpart of umls that have semantic 
> correspondence to the English umls. Topic map is also interesting. I 
> also wonder if there is a direct mapping between topic map and semantic 
> web (rdf/owl) ....
> I agree that we should narrow the scope of our problem a little bit. 
> Otherwise, things tend to fall apart if we try to be too ambitious. I 
> hope we can start thinking more about this Huperzine use case, for 
> example. I also hope such a use case is holistic in the sense that it is 
> both scientifically and technologically interesting.
> Thanks,
> -Kei
> eric neumann wrote:
>> Why not simply use to following trick on top of universal symbols?
>> <umls:male rdfs:label="male" lang="en"
>> rdfs:label="Mann" lang="ge"
>> rdfs:label="mâle" lang="fr"
>> rdfs:label="男性" lang="zh-Hans"
>> ...
>> >
>> Eric
>> 2008/5/28 Jack Park <jack.park@sri.com <mailto:jack.park@sri.com>>:
>>     In cross-language data integration, it may be a simple matter of
>>     using a
>>     multitude of language-scoped labels in an ontology. Another approach
>>     that has been mentioned on this list many moons back by the late Bill
>>     Bugg was that of applying topic maps to the federation of
>>     heterogeneous
>>     resources, including disparate ontologies that don't easily merge, 
>> and
>>     data sets. Bill was referring to some of my work. Topic maps
>>     provide the
>>     ability to apply as many different names to some entity as
>>     necessary for
>>     all participants to successfully locate what they seek. At the same
>>     time, topic maps can federate each entity with external comments,
>>     dialogues (such as this email message), bookmarks (tags) and
>>     relationships with other entities.
>>     Jack
>>     Xiaoshu Wang wrote:
>>     > Huajun Chen@Zhejiang University wrote:
>>     >> Another challenge is cross-language data integration, which is
>>     actually a
>>     >> job that ontology should do.
>>     >>
>>     > I honestly disagree. Ontology is about the semantics of *being*
>>     but that
>>     > of symbols. It doesn't matter if how "gene" is called, named, or
>>     > written. It symbolize the same objective entities. A URI such as
>>     > http://www.example.com is not written in English. It is just a
>>     bunch of
>>     > symbols. Let's not introduce linguistic issues into data
>>     integration,
>>     > which already have a lot of issues.
>>     >
>>     > Xiaoshu Wang
>>     >> Best wishes, huajun
>>     >>
>>     >> -----邮件原件-----
>>     >> 发件人: public-semweb-lifesci-request@w3.org
>>     <mailto:public-semweb-lifesci-request@w3.org>
>>     >> [mailto:public-semweb-lifesci-request@w3.org
>>     <mailto:public-semweb-lifesci-request@w3.org>] 代表 Matthias Samwald
>>     >> 发送时间: 2008年5月26日 21:22
>>     >> 收件人: kc28@email.med.yale.edu
>>     <mailto:kc28@email.med.yale.edu>; Tim Clark
>>     >> 抄送: M. Scott Marshall; public-semweb-lifesci@w3.org
>>     <mailto:public-semweb-lifesci@w3.org>
>>     >> 主题: Re: KB note
>>     >>
>>     >>
>>     >>
>>     >>> Speaking of national boundaries, I wonder if alternative
>>     medicine (e.g.,
>>     >>> herbal
>>     >>> medicine) would also be of interest to this community. For
>>     example,
>>     >>> Huperzine
>>     >>> is a drug derived from the herb Huperzia serrata. I also
>>     wonder if there
>>     >>> are
>>     >>> hypotheses regarding the study of herbs in the possible
>>     treatment of
>>     >>> neurological diseases.
>>     >>>
>>     >> I would also be very motivated to help in this kind of research.
>>     >> Specifically, Huperzine A would be a very interesting use-case
>>     for our
>>     >> developments. It is a herbal compound with a history in folk
>>     medicine and is
>>     >>
>>     >> available OTC in most countries, yet it rivals the 
>> effectiveness of
>>     >> currently leading Alzheimer medications such as Tacrine. It
>>     also has a dual
>>     >> mode of action that does not only involve acetylcholinesterase
>>     inhibition,
>>     >> but also modulation of the NMDA receptor. The implications of
>>     this for the
>>     >> treatment of Alzheimer's are still a rather hot topic.
>>     >>
>>     >> The integration of knowledge from traditional medicine, plant
>>     >> taxonomy/phylogeny/biochemistry and receptor binding databases
>>     (PDSP Ki
>>     >> database, IUPHAR) could lead to the identification of some
>>     extremely novel
>>     >> therapeutic strategies. Finding candidate molecules in such a
>>     way might be
>>     >> much more effective than weeding through libraries of compounds
>>     generated by
>>     >>
>>     >> combinatorial synthesis etc. The challenge lies in the
>>     integration of some
>>     >> very heterogenous datasets that come from vastly different
>>     disciplines,
>>     >> which is exactly the field of research where Semantic Web
>>     technologies are
>>     >> most effective.
>>     >>
>>     >> I guess the major problem for this kind of research is that
>>     there are no
>>     >> funding programmes that span China, the US and Asia...
>>     >>
>>     >> Cheers,
>>     >> Matthias Samwald
>>     >>
>>     >> DERI Galway, Ireland // Semantic Web Company, Austria
>>     >> http://www.deri.ie/
>>     >> http://www.semantic-web.at/
>>     >>
>>     >>
>>     >>
>>     >>
>>     >>
>>     >>
Received on Wednesday, 28 May 2008 21:33:56 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 17:20:34 UTC