MTSR会议�?�续报�?�：三个特邀报告的讨论

MTSR会议（http://mtsr.sigsemis.org/）的网上讨论�?过延长，已于今天（12月5日）结�?�，�?知网站今�?�还能�?能进去，在此想把会议讨论的东西�?总结一下。主�?是三个特邀报告和7个专题的情况，作为11月24日的帖�?“首届元数�?�与语义研究�?网络会议报�?�的补充。由于�?附原文，比较长，这里先是三个特邀报告的一些讨论补充。

一�?
Invited presentation: Dr. Ambjörn Naeve
题目：Some Ideas on Capturing the Semantics of Process Interaction and Process Metadata
报告人：AmbjornNaeve

主�?内容：会议网站�?�给了一个pdf演示稿，�?�能窥其概貌。这是个纯教育领域应用语义Web技术的一个范例，�?点在于建立一个�??分�?�?�的语义模型， �??供以学习者为中心而�?是以教学者为中心的�?把所有知识�?�元打散的�?通过模型�?�散耦�?�的�?“智能化�?的课件�?务（我的�?�解，�?一定正确，因为这样一�?架构 �?实现一定的信�?��?�?，�?�时�??供以学习者为中心的基于问题的学习环境——称为knowledge-pull based on interest），……还有以下特点（抱歉�?�无法翻译了）：

• can beconsidered as a patchwork of knowledge, with a number of linked knowledge patches, each with its own knowledge gardener.
• allows the user to ask questions and search for certified live knowledge sources.
• has accessto a distributed archive of resource components.
• allows teachers to compose components and construct tailor-made learning modules.
• makes use of conceptual modeling in order to support separation between context and content.
• contains aconcept browser (Conzilla) that supports these principles and activities.

简直神了�?�?�说这个系统的需求是基于下�?�的教学�?�设�??出�?�的，实在是很有�?�?：

• Nobody canteach you anything. A good teacher can inspire you to learn.
• Your motivation to learn is based on direct experience of knowledge enthusiasm and positive expectations from live teachers.
• The prerequisites for efficient learning are improved when you take control of your own learning process.
• No deep questions can be answered in an automated way.
• Respect for ignorance – which is of fundamental importance in a non-elitist knowledge society – can only be upheld if the ignorant person is uneducated.

这个系统�?��?� PKLME: PublicKnowledge and Learning Management Environment)

关于讨论：

至今�?�有一个关于作者所开�?�的概念模型系统Conzilla与UML顶层（upper level）的关系问题，希望举一些简�?�的例�?。没有回�?。

二�?
Invited presentation: Dr. Amit P. Sheth
报告人：Dr. AmitP. Sheth，LSDIS Lab, The University of Georgia and Semagix
题 �??：基于语义的�?业和科学应用：研究�?技术与应用实施（Semantics Enabled Industrial and Scientific Applications: Research,Technology and Deployed Applications）

上次（http://keven.wordpress.com/2005/11/24/）介�?了这次会议上�??交的文档和讨论的主�?内容，主�?是语义 Web在目�?�?业信�?�管�?�（KM）的应用，愚以为代表了当今语义Web应用的最高水平，对语义Web应用感兴趣的大虾们切�?�?�错过，仔细看看相关文档，功 力一定大有精进。我也在学习中，这里就�?多说了。

新增的问题讨论：

Murat Osman Unalir问了一个关于“�?�信的�?本体映射问题，�?�如何相信一个映射�?至于产生概念误导？Are there any work on trust aspects ofontology mapping? In other words, can we inject trust into ontology mapping process?

Amit回答说关于�?�信任的语义Web研究很多，但是还没有专门研究本体映射的�?�信任问题的。映射信�?��?�以作为“�?�信任�?务组�?� (composition)�?中的一个专门问题进行研究。并�??供了这一领域研究的一�?教授和项目�??称（Anupam Joshi�?�士的SemDis项目）。
There is a lot of work (incl PhD thesis) on trust in Semantic Web and ontologies, but none that I know of on mapping.
You are right in indicating that just as there is an issue of trust in data or knowledge, a mapping is also a piece of information that by itself could have atrust value/consideration. Some of the work on trust composition could be applicable. You can contact Dr. Anupam Joshi for his thoughts on this matterfor further insights. He is a collaborator on SemDis project. (Google: SemDis,Anupam Joshi)

�?�以说关于元数�?�质�?问题的讨论是本次会议最热门的帖�?，上次贴出一些问答，这里就�?�?�?了。

这个线程里�?�有一个“建立语义Web学院�?的设想获得了热烈的�?应，从12月1日的回�?�?�看�?有�?�能在欧洲建立第一个语义Web学院。大师们谈论了专业方�?�（AI/DB/IR甚至BIS商业信�?�系统）�?学�?�?感兴趣的学校以�?��?�能获得资助的基金等等。

三�?
Invited presentation: Dr. Tom Gruber
Tom的这篇Invited talk的有关资料在这里（如果�?引用，请�?�照下列格�?和出处）：
Thomas Gruber (2005). Folksonomy of Ontology: A Mash-up of Apples and Oranges. First on-Line conference on Metadata and Semantics Research (MTSR’05). http://mtsr.sigsemis.org/. Paper availableat http://tomgruber.org/writing/mtsr05-ontology-of-folksonomy.htm

除了上次报�?�过的Hans的那个问题和Tom的回答之外，还有一些问答。

一个是Hend问的关于Folksonomy与关键�?抽�?�的问题：
I want to ask please, what is the diffirenece between folksonomies and keyword extraction techniques, in other words, when we run an online resource on a keyword extractor program does it give us the same results as what people have taged this resource (i.e.folksonomies)?
Moreover, can we say that each folksonomy word = a keyword = a concept?

Tom没有回答。我的�?�解是folksonomy与关键�?抽�?�没有本质的�?�?�，但是在产生程�?和方�?上�?�能有所�?�?�，folksonomy应该多为 “赋�?�?，�?�以是文中没有（抽�?��?到）的�?，甚至�?�以是一个�?��?（例如分类�?�等），而关键�?抽�?�常常�?�能产生于文本对象中。因而folksonomy�?�以 用于图片�?音视频等多媒体对象。Folksonomy�?一定�?过规范化，因而说folksonomy word=keyword=concept是�?对的，但是他们也是�?�以互相转化的。转化之�?�是�?是应该�?��?�taxonomy，甚至�?过附注概念之间的关系 约�?�并�?过形�?化之�?�能够�?��?ontology，这�??怕需�?业界制定标准�?能解决目�?这些概念的边界问题，从而彻底解决这些新生事物的概念之争。当然这个 想法是个乌托邦。

Ding Yihong（明显是中国人）�??到了一个区分Ontology与Folksonomy的观点：共享的范围�?�?�。Ontology应该能够获得多数人的共识，而Folksonomy�?�是少数人甚至个人的�?表体系。

scale of agreements, folksonomy and ontology

Maybe it is a little late. But I have a question on folksonomy and ontology. Upon my understanding, ontology is a matter of “knowledge sharing”. Then what is the scale of agreement anontology must hold? Must the specifications within an ontology be agreed by everyone on the world? Can it just be upon a small group of people,or even by a person himself? In my mind,I think folksonomy could be viewed as a “personal” ontology; and anontology is just a folksonomy that have been agreed by many people. (although the sophistication of their internal structures could be significantly different.) It only matters of the scale of agreements. I am wondering whether this understanding makes sense or not. Thank you.

就这个问题Tom出�?�澄清了两点：

第一：本体�?�?�以被部分认�?�的，�?一定是一个Ontology中的所有术语，�?�以�?�是一个�?集（这一点多�?我们的元数�?�应用纲�?AP的设计�?�念 啊），�?�以映射到具体的应用系统中去。本体的好�??甚至�?�以用其术语的被接�?�程度�?�评价。第二：tag是个人数�?�，而tags所组�?的folksonomy 则已�?为一组人群所共用的数�?�了，TagOntology（头一次�?�到这个概念，实在是越�?�越混乱了）并�?�是为了在folksonomy的使用中对�?些 tag达�?共识而�??出的，而是为了便于tag标注软件共享tag而使用。我的�?�解是并�?�是基于“概念�?而规范的，而仅仅是为了使用的方便性而�??供一些“�?� 想�?“导航�?功能，犹如全文检索的相关性排�?一样，并�?能�?�到�?�常�?�乎人的需�?。

First, any ontology can be agreed topartially. Ontological commitment is a commitment to use the terms of theontology in a manner that is logically consistent with the defintions. Itis not an agreement to “implement” or have an understanding of any ofthe terms. Typically, a parties that agree to the ontology will use subsets of the ontology (those that map on to the concepts in their systems). Later, a party may choose to have their program handle more terms in the ontology. If the ontology is designed well and in asocial context, then the party will be motivated to reuse the terms from the common ontology instead of making up new, incompatible terms for their system.

The second point is about folksonomy. Tags are personal data, but folksonomies are — by definition — the emergenttag data from a population of users. The point of the TagOntology is NOT to have an agreement about the specific tags used in some folksonomy. Itis to enable tagging software that can share sets of tags and assertions using them.

这个问题紧接�?�就有人（AmitSheth）问了：

你说本体�?�以部分认�?�，那么我们现在是�?�有一�?应用或实现机制，�?�以“部分�?地利用本体呢？一个代�?�如何知�?�一个本体的部分是获得认�?�的，互�?作会�?�生问题�?�？这也是个很好的问题，�?�是对了解AP的元数�?�专家们（比如我们，脸皮厚一点的说）已�?�?�?为问题了。

I am a bit concerned by your comment related to partial agreements related to an ontology. Do we have a mechanismto indicate this partial ontological commitment. What are the interoperability implications of two agents agreeting to and using differentparts of the same ontology?

Tom没有回答而Ding倒写了一篇读�?�感。首先是关于Tom回答她?的问题的，她认为Tom说Ontology的建立完全是基于个人观点的（我没 有看出�?�Tom哪里说了这个�?，Tom好�?指的是Folksonomy）。基于此她认为建立Ontology更应该采�?�Bottom-up或者 Middle-out方法，这样Ontology的“�?��?�?�?更具有实用价值。从实践中�?�考察我也�?��?这个说法，虽然说�?出很好的�?�由。但是我�?�认为 Bottom-up也是需�?在一个domain的群体之间形�?一定的共识，�?能建立domain ontology，而且借鉴top-down的方法能够共享方法论，以�?�更好地达�?本体间的互�?作（也就是接近于ding所说的本体的映射和�?�并等）。

对于Amith进一步问的这个问题，Ding的看法比较�?对，她认为�?达�?两个Agent之间的对�?，必须�?�时接�?�一个Ontology的所有方 �?�，�?�则就�?�?�想象。我还�?知�?�基于Ontology的Agent的行为有哪些实现机制，但仅从元数�?�AP得�?�的�?验，两个Agents在应用中共享 Ontology是完全�?�以�?�到局部共享的，因为Ontology的术语是命�??域独立的，虽然在�?个具体的Ontology中有Context�?�系，脱离 了整体Onto会�?�失Context，然而�?�以在应用系统中采用继承�?�?定等方�?产生术语间新的Context，这样的Ontology的应用 （runtime）也是关系�?�计算的，形�?化的。

After reading the comments from both ofyou, I have some thinking, please correct me when I get wrong.

From Tom’s answers for my question, itlooks like that the construction of ontology reflects only personal view forthe domain of interest. Other parties may (or may not) agree on thespecifications in an ontology. This understanding actually makes senses to mebecause I believe that since the Web itself is an open world, everyone canpublish his view to the world, and every other person can simply agree to orreject the other person’s points. When more and more people agree on and use,for example, the specifications in one ontology, it becomes a standard withinthe social context. If this is thepicture, the formalization of ontologies must be a bottom-up (or middle-out)process. In other words, it is some-how conflict with the current populartop-down ontology-constrcution paradigm. Nevertheless, the formalization of the leaf-level semantics could bemore important than the construction of the internal structure of ontologiesbecause the former one may help lead to advanced ontology mapping and mergingmechanisms.

I am not trying to replace Dr. Gruber toanswer questions. But I would like to present my thinking of Dr. Sheth’s quesition. If two agents agree to and use different parts of the same ontology,and they want to communicate with each other, they may have to accept the part of the ontology they previously disagree with. It likes the communication amonghuman beings, one person may not agree on the other person’s terms. He,therefore, can simply igore the one he doesn’t like and refuse to talk withhim. But if the first one expect to communicate with the second guy, he has to compromise the situation and agree on using the terms the second one uses. I am very interested in knowing what Tom’s response to this question.

总的�?�说，这次网络会议在组织形�?�?技术平�?��?�?�加人数�?讨论进展等方�?�并�?能尽如人�?，我在Metadata intensive applications and Semantic Web Technology讨论组中�??出的问题并上载了pdf和ppt文件�?�有很少人访问，而且居然没有人�??供�?�?，让人大跌眼镜。但是�?管怎么说，通过交�?， 知�?�了这个领域的牛人们在�?�些什么，�?��?了自己的一些�?�解和想法是�?�有根�?�，是�?�有根本性的错误或�??差，也算收获很大。国内�?�Semantic Web的人应该说�?少，但是�?�自为政，交�?很少，也难以使项目和技术�?果有一顶的系统性和延续性，甚至在我看�?�国内有�?少计算机的硕士�?�士对 Ontology�?�相关semantic web技术存在一定认识上的�??差，或者过于注�?从计算机科学基础（如算法�?�逻辑）方�?�进行�?明或求得�?破与进展，对于应用技术领域，如体系架构�?应用模 型�?标准规范等方�?�比较轻视，也因此国内有关技术的应用实践方�?�基本上是一片空白，为国外培养了大批人�?，将�?�还�?花大钱应用国外的技术。

（七个专题讨论的�?�独介�?�?��?下期）