|
Transcript of the KB online event with the REWERSE project team, taken from the KB online workshop space, 17 May 2005 Ron Dvir: Hello all, I am most happy to welcome you to the REWERSE on line event. Can I ask each to introduce him/herself - one sentence?
My name is Ron Dvir, expert in Innovation Management and Intellectual Capital. Looking to learn about your subject!
Mounib Mekhilef: Mounib, I'm here as the Community Steward of the KnowledgeBoard and also as a KM practicioner from the "hard" side. looking to learn more...
Nicola Henze: Hi, I'm Nicola Henze. I am doing research on personalized applications for the Semantic Web.
Uta Schwertel: Hello all, my name is Uta Schwertel. I am here as the project manager of the project REWERSE.
François Bry: My name is François Bry, from Munich. I work on Web and Semantic Web query language -- and contribute to REWERSE's management.
Ron Dvir: Massimo, who are you?
Andrea Kulas: Hi, I'm Andrea Kulas. I am responsible for technology transfer for REWERSE.
Massimo: Always in a hurry... but thanks for asking Ron ;)
Ron Dvir: Ahha, so you are the great singer from Bologna, that Massimo?
Massimo: I wish ;) Anyway, I am Massimo Marchiori, working in W3C/MIT and the University of Venice, exec comm member of REWERSE
Ron Dvir: You wrote that "REWERSE is a Network of Excellence on Reasoning on the Web with Rules and SEmantics". What does "REASONING on the Web" this mean - pls. explain to us, those guys who are not from the Semantics field.
François Bry: Massimo, will you be at the REWERSE I3-I4 meeting in Venice on June 20-23? I hope, you will make it to Venice :-)
François Bry: The answer is given the well-known Scientific American paper of 2001 by Tim Berners-Lee, James Hendler, and Ora Lassila: ``For the semantic web to function, computers must have access to [...] sets of inference rules that they can use to conduct automated reasoning.''
François Bry: Another answer to the question "Why does one need reasoning on the Web?" is as follows. Data on the Web, and on the Semantic Web, are retrieved using declarative query languages, i.e. inspired from logic and functional programming. Queries are like "database views", i.e. if-then formulas. Combining queries, are they pre-defined or ad hoc, is therefore like combining, or chaining, if-then formulas, i.e. is a simple form of reasoning -- in fact reasoning in a constructive logic.
Massimo: Francois, actually I am losing much time in trying to organize the rooms, meals, so I hope to be there, yes ;)
François Bry: A further answer to the question "Why does one need reasoning on the Web?" is that the Semantic Web builds upon logic formulas in various formalisms e.g. RDF and OWL. Extracting the substance from such specification is reasoning.
Ron Dvir: going for the basics for a sec - give an example of reasoning over the web - and how semantics get into the picture
François Bry: You look for a pharmacy in your neighbourhood. Obviously, you want a pharmacy which currently is open. Further, you want to easily reach it, depending on whether you walk, drive a bike or a car. If you drive a car, you want to park it close to the pharmacy. Coping with all this is reasoning - what one does without thinking much of it and what current sweb software do not do much.
Eddy Kloprogge: Hi there
Mounib Mekhilef: so is there a difference between this an what have been studied during several decades through "expert system"
Ron Dvir: Francois, and jst to conclude your example - what is the semantics in the case of the pharmacy example
François Bry: Another example is when one look for a cinema where, say, Hotel Rwanda is played. Depending on one profile, the cinema might have to accept wheelchairs, the movie might be preferably in English, and close to a vegetarian restaurant. Coping with such wishes and combine them in a consistent and sensible manner is reasoning.
Uta Schwertel: An example would be someone organising a trip over the net. She books a flight and a hotel. If e.g the flight is re-scheduled a re-booking of the hotel is necessary. To do this re-booking automatically reasoning is necessary, e.g. about time, availability of hotel rooms etc.
Ron Dvir: And back to mounib question - is this what we used to call Artificial Intelligence?
Nicola Henze: Yes, there is an important difference to "expert systems": an expert system is designed by some people, and the information in the expert system is under control (update, modifications, etc.). In the Semantic Web, there is no such "controlled condition": you have dynamic and changing knowledge base in which you want to perform - indeed similar - tasks.
Maya Levin-Sagi: sorry, have had tech problems, this is Maya from Edna Pasher, am a social psychologist
Mounib Mekhilef: welcome
François Bry: Semantics is a technical word. I am not sure it is good to start with technical stuff. Semantics is Greek for meaning. My guess is that almost everyone would grasp the "meaning" of this example. If now you formalize this example using declarative, or logic languages, then you might -- but not necessarily need- - to refer, say to proof systems and their correctness. This is where "semantics" gets in. But rthis is "dev3eloppers' stuff.
Massimo: it is in fact true there is confusion between the
Mounib Mekhilef: back to your justification. A part from the expertise, is the process to infere the rules the same?
Ron Dvir: OK, now we (alt least stupid me) understand from the examples what is reasoning. What is YOUR big challenge? What is missing from current solutions?
Massimo: "broad" semantic web interpretation ("gives intelligence to the web") and the more technical term, which is a standardized way of accessing information on the web
François Bry: Massimo, hold on the line!
Massimo: sorry, I have refresh problems (must be Mozilla's)
François Bry: What is missing from current solutions? (1) they are not widespread. (2) they are not easy to implement. (3) They are incomplete. (4) Not so many people understand them.
Ron Dvir: hi Gerd, welcome! who are you?
Gerd Wagner: I'm from REWERSE!
François Bry: Massimo> sorry, I have refresh problems. Take a drink! :-)
Ron Dvir: Francois mentioned 4 things which are missing from current solutions. WOW, is it such virgin field, or is the complexity so big?
Massimo: See, that's another perfect application of potential Semantic Web techniques.... someone enters this chat and you would like to know who s/he is...
Massimo: ... and what his relationship to the SW interests and REWERSE are...
François Bry: Indeed, reasoning is a rather virgin field. It started only a few century before Christ (in Greece), it has been formalized as a computation only in the 17th century, and its automation has only started a few decades ago. That is, almost everything remains to be done. We work on that.
Ron Dvir: Guys, so what is your Dream? I mean, when the project REWERSE is finished, what will the world have? Where are you planning your breakthroughs?
Nicola Henze: What is missing from current solutions is, that often the information that you are looking for is somewhere out there in the Web. But we lack efficient and effective support from applications that allow us to extract or combine or rate this information, or to detect the context and the consequences, etc.
François Bry: when the project REWERSE is finished, what will the world have? A lot of research proposals, papers, prototypes, use cases, test beds. Our dream is to do decent work on this. I do not understand the breakthrough question. Reasoning is the breakthrough -- for humanity as well as for computers. This breakthrough goes on.
Ron Dvir: Nicola, you mention a classical KM problem: "the inofrmation is there, but difficult to extract value out of it". what are you doing about it?
Maya Levin-Sagi: how does this compare to search engines that already exist out there? for example Google? does this have anything to do with them?
Uta Schwertel: We are working on a set of languages and techniques that move us a step forward to model reasoning over the Web automatically. This will be one step in a more user-friendly and more efficient Web.
François Bry: Search engines provides with indexing. They surely apply some form of reasoning for building indexes.
Nicola Henze: For example, we have developed some "watchers" on dedicated Web sites which regularly extract the information from those sites and transcript it into Semantic Web languages. This information can now be syndicated according to users preferences.
Ron Dvir: I just looked at your nice blue diagram on the event page - the "pyramid" of trust, proof, logic, ontology, rdf etc. When are you focusing on this scheme? or is it an holistic approach?
Massimo: The search engines of the future will surely apply some form of semantic web reasoning, the potential is just too big. They are already exploding into something more (see eg Google Maps, or Yahoo's CC service)
François Bry: REWERSE works on the logic and reasoning layers.
Mounib Mekhilef: to what extent the techniques you are developing are language and culture independent?
François Bry: To what extent the techniques you are developing are language and culture independent? Do you mean "natural language" or "programming language"?
Maya Levin-Sagi: great question, Mounib!
Mounib Mekhilef: natural language
François Bry: Well, this is an interesting point.
Ron Dvir: and...
Uta Schwertel: Since we use formal representations I think the *techniques* are not language specific. Only if you have natural language interfaces to the formal representations language becomes relevant.
Mounib Mekhilef: I said this cause I remember I had trouble during a long time dealing with yahoo and Google, while friends success finding stuff I always miss the essential. I was formalizing too much
Ron Dvir: ahha, mounib comment make me think - is the best strategy to go for strictly formal rules, or perhaps more loose rules and semantic network, which can "respond" to different situations?
François Bry: Well, this is an interesting point. Logic, as irt is know since the old Greeks is highly "natural language" dependent. It is based on the one of the two families of natural languages humanity has developed, the one building upon triples of the form (subject, property, object). The second family of natural languages builds upon "contexts", has no such triples. I am dreaming of investigating Knowledge Representation using a framework inspired from the "context" approach to natural languages. Maybe after REWERSE, maybe during a long stay in China, Korea or Japan.
Maya Levin-Sagi: It does seem that semantics of someone from France could be different than that of someone from Russia or Morocco
François Bry: Indeed, this is one of the point that the Pharmacy example I previously gave stresses.
Mounib Mekhilef: I would say that there is another "problem" The way we make our reasoning is different if we come from social science (as an example) and Math in addition to what you mentioned maya
François Bry: REWERSE investigates cultural aspects, too. A REWERSE member had a paper at WWW 2005 last week in Chiba, c lose to Tokyo. The paper was about modelling professional and/or cultural calendars.
François Bry: The way we make our reasoning is different if we come from ...
Ron Dvir: what is a "cultural calendar" (fascinating term) - and how web reasoning is related to it
François Bry: I am not sure. I basic, often implicit but nonetheless important assumption is that reasoning is common to all mankind. There is eg no "female reasoning" not accessible to males. Indeed, they might be preferences, predilections, or predispositions. But everyone can share the "reasoning" of everyone else.
Uta Schwertel: A cultural calendar contains e.g. the specific religious holidays of a country. A Hebrew calendar is different from a Christian calendar as to these holidays.
François Bry: "Cultural calendar" refers to calendars like the one used in Germany (Gregorian with 2 CXmas days), in Greece (Gregorian, too, but with the date of Eastern determined from a revised Julian calendar and the Easter time lasting for a full wee), the Hebrew calendar, the Japanese calendar, etc. They are all much more different than one might think at first. Tackling all these calendars might be the premier challenge of "Web internationalization - a challenge REWERSE is working on.
Ron Dvir: Francois says - "But everyone can share the "reasoning" of everyone else"!!!! That is such a strong statement . If this was completely true, would we still have conflicts? (sorry, this is slightly a side track...)
Mounib Mekhilef: Another question: People working in semantic web, talk also about ontology (see for ex the works of Rose Dieng from Sophia Antipolis). what can we learn from this?
François Bry: Francois says - "But everyone can share the "reasoning" of everyone else"!!!! That is such a strong statement . If this was completely true, would we still have conflicts? (sorry, this is slightly a side track...)
François Bry: The fact that we can does not mean that we do. Obviously we do and we don’t. E.g. French and German are quite reasonable these days, Israeli and Palestinian are not.
Ron Dvir: François - this makes sense: we have the capacity to share reasoning, but sometimes we don’t practice this human potential
Ron Dvir: What kind of Semantic Web applications that are not (or not as much) considered in other Semantic Web projects is REWERSE striving to make possible?
François Bry: REWERSE has a first bias towards combining standard forms of reasoning, the well-established Description Logics being considered as standard, and non-standard reasoning forms. For example, if one wants to schedule a conference with participants distributed over the world, then one needs computing time shifts because of different time zones and solving constraint problems. For this, one cannot rely on standard forms of reasoning. One needs a good deal of (simple!) arithmetic and constraint reasoning.
François Bry: REWERSE has a second bias towards Bioinformatics. Bioinformatics is a great "playground" or "Sandkasten" for the Semantic Web. As e.g. stressed in the Book "Data on the Web" by Serge Abiteboul, Peter Buneman, and Dan Suciu (1999), one of the first real-life application and occurrence of what has been later called "semi-structured data" was in Bioinformatics. In the article "A Computational Biology Database Digest: Data, Data Analysis, and Data Management" (Distributed and Parallel Databases 13(1): 7-42, 2003), Peer Kröger and myself show how web query languages dedicated to Bioinformatics data bases introduced Web-specific query facilities long before they have been considered in the context of general purpose Web query languages. Bioinformatics gives REWERSE a concrete and highly productive anchorage in an extremely dynamic application field for all kind of Web and Semantic web techniques.
François Bry: REWERSE has a third bias towards "personalized" Web-based systems, i.e. systems such as eCommerce recommender, eLerning systems and the like. Personalization is considered in REWERSE to be one of the field of today's applied Computer Science which will gain in practical and economical importance. Most likely, personalization will widespread so much that nobody will be conscious that this is used to be a specialized area in applied Computer Science.
François Bry: A further aspect that might be quite specific to REWERSE is its striving towards interoperable and coherent languages and formalisms. In simple words, this means: as few reasoning languages as possible, that are as complementary as possible, and that are as similar as possible. In a way, this striving of REWERSE for minimality, complementarity, and similarity makes REWERSE very special in today's Web and Semantic Web research where there similar yet different formalisms are burgeoning in all possible manners. Note that REWERSE has no proviso at all against this burgeoning, quite the opposite: it is useful and maybe necessary. But it leaves room for another approach, one trying to "bring things, ideas, and people together".
Nicola Henze: REWERSE is focussing on the formal languages that are required for performing reasoning tasks on the Web. This includes having for example rules that can be less stricted (rule relaxation) under specific circumstances, or defeasible rule systems, etc.
Mounib Mekhilef: Do you consider "fuzzy reasoning"? to build less restricted rules?
Ron Dvir: Wow, REWERSE has clear conceptual approach (which you called "bias"). Can you tell as the story of one of your test beds and how these 3-4 "bias" are realized in them (e.g. how rules are relaxed etc.)
François Bry: "fuzzy reasoning" is considered, too. Hans Jürgen Ohlbach has designed and implement very nice algorithm for coping with fuzzy time intervals or locations as one need for coping with questions like "a pharmacy in the South of Munich which is open in the early morning". Think of it, answering such question automatically in an acceptable manner is a great challenge -- not solved today but well investigated in REWERSE.
Mounib Mekhilef: Is there any active example (prototype) that we can play with?
Nicola Henze: It must not be fuzzy reasoning. For example, if you have a rule "if apples then ..." but you cannot proof "apples" but only "fruit". And, from a conceptual description(ontology) you can derive that apples are fruit, the relaxed rule can be considered (if the context allows for relaxation).
Mounib Mekhilef: get it!
François Bry: Michael Schroeder, who unfortunately could not join today, has shown how querying a Gene molecular Biology database is very conveniently, ie effectively, done using rules. We re-implemented this using REWERSE's own Web query language, what came up with very effective queriores, i.e. queries that are rather easy to read, understand, and hopefully right. I do not remember by heart the references of the paper. Look at the Web site http://www.pms.ifi.lmu.de/publikationen/. The paper is there.
François Bry: A Web query language prototype is accessible at http://xcerpt.org. See also http://rewerse.net. But do not expect too much today: REWERSE is only 1 year old and made up of mostly university people, i.e. people that have to do too many different things at one!
Nicola Henze: You can also play with a personal publication browser (available via www.personal-reader.de) which shows scientific publications of the REWERSE project in the context of this project. Beware: you won't see the rules behind the applications but they are there and reason about distributed data :-)
Ron Dvir: What type of reasoning languages does the Web, especially the Semantic Web need?
François Bry: In my opinion, the Web and especially the Semantic Web need logic, or reasoning, languages of the following different kinds (cf. "Ten Theses on Logic Languages for the Semantic Web" François Bry and Massimo Marchiori, Proc. W3C Workshop on Rule Languages for Interoperability, April 2005),) : (1) three kinds of reasoning, or deductive, languages, namely (1.1) constructive rules (or views), (1.2) normative rules (or integrity constraints), (1.3) descriptive specifications (or ontologies), and another fully different kind of languages, namely (2) reactive rules. Note that "reactive rules" like "production rules" and "Event-Condition-Action rules" are an *imperative* languages in the disguise of declarative rules. Indeed, they spec ify state changes.
François Bry: Missing in the above list are the reasoning languages needed for policy, e.g. trust, specifications and negotiation. As of today, it is rather unclear what reasoning languages will be needed for policy and negociation. My own impression is that declarative languages from the above-mentioned types, viz. constructive rules (or views), normative rules (or integrity constraints) and descriptive specifications (or ontologies) will most likely be sufficient for specifying policies while "negotiation reasoning" will call for novel reasoning paradigms similar to those investigated during the last decade for formalizing arguments, argumentation semantics, and the like e.g. Michael Schroeder, Ralf Schweimeier: Arguments and Misunderstandings: Fuzzy Unification for Negotiating Agents. CLIMA III 2002: 1-18.
Mounib Mekhilef: thanks. I just made a test I wrote (In French! ) "Livre Jules Verne" which supposed that I'm looking for books written by Jules Verne. the result I got from google is not consistent since at the third level I have a TV show close to the topic but not so much. What REWERSE can do, what does it offer. sorry for the naivety!
François Bry: That's the typical SW test question, isn't it? One and half year ago, amazon.com offered me trav el guides to Greece as they could not find Grigoris Antoniou's book on the SW. Avoiding this calls for annotating data with meta-data. REWERSE will provide tools easing querying data and meta-data. Note, a salient specificity of REWEWRSE's query language is its "versatility": one can access both data eg in HTML or XML and meta-data in RDF or OWL in a same query and the QL offers reasoning. I know of no other Web or SW QL doing this.
Ron Dvir: This is the
Ron Dvir: This is almost the last question - what is your biggest challenge, I mean what is this "key issue" or secret that if you solve you will get much better solution?
François Bry: I believe, our biggest challenge is to do decent research, to work together, and to contribute to sketch a vision for the years after the REWERSE project. REWERSE is a "Network of Excellence" ie it primarily aims at bringing people and idea together. This indeed is a great challenge. Maybe the greatest one might think of.
François Bry: Another, more technical and more immediate challenge are the various kinds of reasoning the SW will need.
François Bry: My conviction is that the different kind of languages the Semantic Web needs (in Massimo's and my view constructive rules (or views), normative rules (or integrity constraints) and descriptive specifications (or ontologies) as well as reactive rules) call for *different* forms of reasoning, especially different treatment of negation - this is developed in the formerly mentioned paper "Ten Theses on Logic Languages for the Semantic Web".
Gerd Wagner: I don't think there is such a "key issue" which is the key to everything. We have many key issues but no magic!
François Bry: You see, I do not loose a chance to advertise for REWERSE papers.
Ron Dvir: got it, no magic, no tricks, lots os hard serious research, step after step to frogress in this complicate domain. Good luck!
Ron Dvir: We are about to conclude this fast "ping pong" dialogue - would everybody want to make one concluding comment (e.g. what do you take from this conversation)?
François Bry: Thanks for the wishes!
Mounib Mekhilef: thanks for all. I would invite you to continue this discussion offline on the KnowledgeBoard.
François Bry: It was interesting to hear how each of us from REWERSE address the REWERSE goals and work, and to hear how colleagues from outside REWERSE look at its work.
François Bry: Bye!
Uta Schwertel: One of the "key issues" of REWERSE is also to disseminate its ideas and technologies. This was interesting platform to talk to people outside REWERSE about our ideas. Thanks!
Mounib Mekhilef: thanks for all and for REWERSE team to share with us.
Maya Levin-Sagi: Good luck! Bye!
Nicola Henze: Thanks to all! I think the questions show that there is big interest and need to improve our current "Web" for a more supportive Semantic Web.
Massimo: In fact, the interesting part is to see the kind of questions that have been posed here, in that the majority of them asked for concrete tools, solutions, and the "magic bullet"...
Gerd Wagner: Bye
Nicola Henze: Bye
Mounib Mekhilef: bye
please can someone save the transcript
Andrea Kulas: bye
Massimo: Which shows that really the next challenge of the Semantic Web is to connect to the people... (cf http://www.w3.org/RDF/Metalog/docs/sw-easy )
Massimo: bye everyone
Ron Dvir: thank you all, for me it was interesting to learn about this domain - and I am sure that the transcript will be useful for many additional members of KnowledgeBoard. It will be posted in about a week, and then you might want also to post it in your site? Anyway, thanks all!!!
|
