From OntologPSMW

Jump to: navigation, search
[ ]

Contents

RulesReasoningLP: mini-series session-03 - Thu 2013-11-21     (1)

Program: Ontology, Rules, and Logic Programming for Reasoning and Applications (RulesReasoningLP) mini-series of virtual panel sessions     (1A)

Topic: Concepts and Foundations of Rules and Ontologies: Logic Programs, Classical Logic, and Semantic Web - II     (1B)

Session Co-chairs: Dr. Leo Obrst (Ontolog; MITRE) & Professor Pascal Hitzler (Wright State U) ... intro slides     (1C)

Panelists / Briefings:     (1D)

  • Dial-in:     (1J4)
    • Phone (US): +1 (206) 402-0100 ... (long distance cost may apply)     (1J4A)
    • in view of recently reported skype connection issues, this is not recommended (especially for speakers) although it may still work for some ... Skype: joinconference (i.e. make a skype call to the contact with skypeID="joinconference") ... (generally free-of-charge, when connecting from your computer ... ref.)     (1J4B)
      • when prompted enter Conference ID: 141184#     (1J4B1)
      • Unfamiliar with how to do this on Skype? ...     (1J4B2)
        • Add the contact "joinconference" to your skype contact list first. To participate in the teleconference, make a skype call to "joinconference", then open the dial pad (see platform-specific instructions below) and enter the Conference ID: 141184# when prompted.     (1J4B2A)
      • Can't find Skype Dial pad? ...     (1J4B3)
        • for Windows Skype users: Can't find Skype Dial pad? ... it's under the "Call" dropdown menu as "Show Dial pad"     (1J4B3A)
        • for Linux Skype users: please note that the dial-pad is only available on v4.1 (or later; or on the earlier Skype versions 2.x,) if the dialpad button is not shown in the call window you need to press the "d" hotkey to enable it. ... (ref.)     (1J4B3B)
      • if you are using skype and the connection to "joinconference" is not holding up, try using (your favorite POTS or VoIP line, etc.) either your phone, skype-out or google-voice and call the US dial-in number: +1 (206) 402-0100 ... when prompted enter Conference ID: 141184#     (1J4B4)
  • Shared-screen support (VNC session), if applicable, will be started 5 minutes before the call at: http://vnc2.cim3.net:5800/     (1J5)
    • view-only password: "ontolog"     (1J5A)
    • if you plan to be logging into this shared-screen option (which the speaker may be navigating), and you are not familiar with the process, please try to call in 5 minutes before the start of the session so that we can work out the connection logistics. Help on this will generally not be available once the presentation starts.     (1J5B)
    • people behind corporate firewalls may have difficulty accessing this. If that is the case, please download the slides above (where applicable) and running them locally. The speaker(s) will prompt you to advance the slides during the talk.     (1J5C)
    • instructions: once you got access to the page, click on the "settings" button, and identify yourself (by modifying the Name field from "anonymous" to your real name, like "JaneDoe").     (1J6A)
    • You can indicate that you want to ask a question verbally by clicking on the "hand" button, and wait for the moderator to call on you; or, type and send your question into the chat window at the bottom of the screen.     (1J6B)
    • thanks to the soaphub.org folks, one can now use a jabber/xmpp client (e.g. gtalk) to join this chatroom. Just add the room as a buddy - (in our case here) ontolog_20131121@soaphub.org ... Handy for mobile devices!     (1J6C)
  • Discussions and Q & A:     (1J7)
    • Nominally, when a presentation is in progress, the moderator will mute everyone, except for the speaker.     (1J7A)
    • To un-mute, press "*7" ... To mute, press "*6" (please mute your phone, especially if you are in a noisy surrounding, or if you are introducing noise, echoes, etc. into the conference line.)     (1J7B)
    • we will usually save all questions and discussions till after all presentations are through. You are encouraged to jot down questions onto the chat-area in the mean time (that way, they get documented; and you might even get some answers in the interim, through the chat.)     (1J7C)
    • During the Q&A / discussion segment (when everyone is muted), If you want to speak or have questions or remarks to make, please raise your hand (virtually) by clicking on the "hand button" (lower right) on the chat session page. You may speak when acknowledged by the session moderator (again, press "*7" on your phone to un-mute). Test your voice and introduce yourself first before proceeding with your remarks, please. (Please remember to click on the "hand button" again (to lower your hand) and press "*6" on your phone to mute yourself after you are done speaking.)     (1J7D)
  • RSVP to peter.yim@cim3.com appreciated, ... or simply just by adding yourself to the "Expected Attendee" list below (if you are a member of the team.)     (1J9)

Attendees     (1K)

  • Expecting:     (1K2)
    • (please add yourself to the list above if you are a member of the community, or, rsvp to <peter.yim@cim3.com> with the event title/date and your name and affiliation)     (1K2B)

Abstract     (1L)

Concepts and Foundations of Rules and Ontologies: Logic Programs, Classical Logic, and Semantic Web - II ... intro slides     (1L1)

This is the 3rd session of the RulesReasoningLP mini-series - a series of virtual panel sessions, and the associated online discourse, co-championed by some members of the Ontolog community who value the importance of the subject matter and would want to bring together those who are knowledgeable or interested into a dialog. The mini-series program will cover the topics that encapsulates the ontology-driven applications that will generally fall under "Ontology, Rules, and Logic Programming for Reasoning and Applications."     (1L2)

This session is the second of two sessions devoted to addressing the concepts and foundations of the technologies underlying ontology and rule reasoning, especially focused on logic programming and Semantic Web extensions. Panelists invited to share their work with us today include Dr. Markus Kroetzsch, Dr. HectorPerezUrbina, Professor HassanAitKaci and Professor EnricoFranconi.     (1L3)

After the panelists briefings, there will be time for Q&A and an open discussion among the panel and all the participants.     (1L4)

See more details at: RulesReasoningLP (homepage for this mini-series)     (1L5)

Briefings     (1L6)

  • Dr. MarkusKroetzsch (Technische Universit��t Dresden) - "Existential Rules in Ontological Modelling" - slides     (1L6A)
    • Abstract: We are witnessing a renaissance of rule-based approaches both in knowledge representation and in databases. Data management applications consider Datalog as a query language while tuple-generating dependencies (TGDs) are used in data integration and exchange. In ontological modelling, existential rules (a.k.a. Datalog+/-) are investigated, leading to many new rule languages of varying expressivity and complexity. In this talk, we give a brief overview of these recent developments, and we present a particular application of rules in knowledge modelling that is contrasted to traditional uses of rules in AI applications.     (1L6A1)
  • Dr. HectorPerezUrbina (Clark & Parsia, LLC) - "Modeling with Rules in Practice" - slides     (1L6B)
    • Abstract: In our experience, users typically regard rules as a friendlier alternative to axioms; however, in spite of their popularity, rules remain poorly understood. In this talk, we review common concerns, questions, and difficulties from our user base regarding the use of rules for modeling ontologies. Such issues range from the semantics of rules and modeling best practices to the use of existing tools. We believe their input is valuable as it provides guidance for creating effective educational materials, directing applied research, and improving and developing tools.     (1L6B1)
  • Professor HassanAitKaci (Universit�� Claude Bernard Lyon 1) - "Reasoning and the Semantic Web" - slides     (1L6C)
    • Abstract: Arguably, what made J.A. Robinson's resolution-based reasoning (and thus Prolog-style Logic Programming) powerful is first-order term (FOT) unification. Unification defines a partial-order on the algebra of FOTs, endowing it with a lattice-structure as shown by G. Plotkin in 1970. Intuitively, a FOT may be construed as a data structure denoting the set of all its instances. Then, unifying two FOTs computes the intersection of their denotations. FOT unification seen as a constraint-solving process consists in a very efficient (linear-time) equation-solving procedure. As such, a FOT is a handy data structure that Prolog programmers use to their benefit for representing approximations of objects. Yet, FOTs can be made much more expressive when seen as Order-Sorted Featured (OSF) graphs. Nodes can be labeled with partially-ordered symbols denoting taxonomic concepts rather than just functional constructors, and subterms can be indexed by unbounded feature symbols rather than fixed-arity positions. Logical variables then become coreference tags expressing equality constraints among feature compositions, including cyclic ones. In this way, all the nice features of FOTs are preserved (e.g., set-of-instance denotation, linear unification). As a result, logic programmers can enjoy a versatile data structure that is a natural extension of, and as efficient as, a Prolog term, with the added bonus of enabling taxonomic reasoning based on OSF graph unification. This presentation will overview and discuss the potential that reasoning using OSF graph unification opens for the Semantic Web thanks to the Constraint Logic Programming paradigm, as opposed to Description Logic tableaux-based reasoning adopted by the W3C. ... (latest version of my slides for this talk is also available here.)     (1L6C1)
  • Professor EnricoFranconi (Free University of Bozen-Bolzano) - "The Logic of Extensional RDFS" - slides     (1L6D)
    • Abstract: The W3C normative version of RDF Schema (RDFS) gives non-standard (intensional) interpretations to some common notions such as classes and properties, thus departing from set-based semantics (such as the OWL semantics) and leading to an unexpected behaviour. In this work we assign the correct set-based (extensional) semantics for the RDFS vocabulary while preserving the simplicity and computational complexity of deduction of the intensional version. This result can positively impact current implementations, as reasoning in RDFS can be implemented as a simple extension to current RDFS engines.     (1L6D1)

Agenda     (1M)

RulesReasoningLP Mini-series Panel Session-03     (1M1)

Session Format: this is a virtual session conducted over an augmented conference call     (1M2)

Proceedings     (1N)

Please refer to the above     (1N1)

IM Chat Transcript captured during the session    (1N2)

see raw transcript here.     (1N2A)

(for better clarity, the version below is a re-organized and lightly edited chat-transcript.)     (1N2B)

Participants are welcome to make light edits to their own contributions as they see fit.     (1N2C)

-- begin in-session chat-transcript --     (1N2D)


Chat transcript from room: ontolog_20131121     (1N2E)

2013-11-21 GMT-08:00 [PST]     (1N2F)


[9:03] Peter P. Yim: Welcome to the     (1N2G)

RulesReasoningLP: mini-series session-03 - Thu 2013-11-21     (1N2H)

Program: Ontology, Rules, and Logic Programming for Reasoning and Applications (RulesReasoningLP)     (1N2I)

Mini-series of virtual panel sessions     (1N2J)

Topic: Concepts and Foundations of Rules and Ontologies: Logic Programs, Classical Logic, and Semantic Web - II     (1N2K)

Session Co-chairs: Dr. Leo Obrst (Ontolog; MITRE) & Professor Pascal Hitzler (Wright State U)     (1N2L)

Panelists / Briefings:     (1N2M)

Logistics:     (1N2R)

  • (if you haven't already done so) please click on "settings" (top center) and morph from "anonymous" to your RealName     (1N2T)
  • Mute control (phone keypad): *7 to un-mute ... *6 to mute     (1N2U)

(i.e. even if it says it is "offline," you should still be able to connect to it.)     (1N2X)

    • if you are using skype and the connection to "joinconference" is not holding up, try using (your favorite POTS or     (1N2Y1)

VoIP line, etc.) either your phone, skype-out or google-voice and call the US dial-in number: +1 (206) 402-0100     (1N2Z)

... when prompted enter Conference ID: 141184#     (1N2AA)

missing on Linux-based Skype v4.x for skype-calls.)     (1N2AE)

Proceedings     (1N2AL)

[9:16] anonymous morphed into Brandon Whitehead     (1N2AM)

[9:23] anonymous1 morphed into HectorPerezUrbina     (1N2AN)

[9:23] anonymous1 morphed into Enrico Franconi     (1N2AO)

[9:26] anonymous morphed into Enrico Franconi     (1N2AP)

[9:29] anonymous morphed into Markus Kroetzsch     (1N2AQ)

[9:30] Brandon Whitehead: [in response to PeterYim's appreciation that Brandon Whitehead is joining us from     (1N2AR)

New Zealand, at a very inconvenient hour] Thanks Peter! It's better now...I much prefer 0630 to 0530. :)     (1N2AS)

[9:30] anonymous morphed into ElieAbiLahoud     (1N2AT)

[9:33] anonymous morphed into Pascal Hitzler     (1N2AU)

[9:35] Hector Perez-Urbina morphed into HectorPerezUrbina     (1N2AV)

[9:36] Pascal Hitzler: Apologies I'm having technical issues with the connection     (1N2AW)

[9:40] Pascal Hitzler: let me try to reconnect     (1N2AX)

[9:41] Pascal Hitzler: perhaps to avoid further loss of time     (1N2AY)

[9:42] Pascal Hitzler: Leo can make an intro     (1N2AZ)

[9:42] Pascal Hitzler: and you can start?     (1N2AAA)

[9:42] anonymous morphed into Conrad Bock     (1N2AAB)

[9:43] Peter P. Yim: == Leo Obrst & Pascal Hitzler starts the session - see slides under     (1N2AAC)

[9:46] anonymous morphed into Onno Paap     (1N2AAE)

[9:48] anonymous morphed into Chuck Rehberg     (1N2AAG)

[9:51] Peter P. Yim: @Leo @Pascal - we have 37 people on the voice bridge, but only 26 in the chat-room     (1N2AAH)

now; so, at the next opportunity (transition to next speaker,) please prompt people to join us in     (1N2AAI)

the chat room     (1N2AAJ)

[9:54] anonymous morphed into David Mendes     (1N2AAK)

[9:54] Amanda Vizedom: @MarkusKroetzsch -- In ontological modeling of complex structures and domains,     (1N2AAL)

a classic approach (in AI applications but not DL traditions, of course) is to model many of the     (1N2AAM)

relatively stable complex relationships between things *as rules*. To me, it seems that removing     (1N2AAN)

rules *as a means of representing complex relationships* from the ontological toolkit makes ontology     (1N2AAO)

work much harder and less effective. So, to me, these views seem to overlap. Do you not agree? Or do     (1N2AAP)

you think that the complex relationships that exist are not to be ontologically (and declaratively)     (1N2AAQ)

modeled, or should be modeled in some other way?     (1N2AAR)

[9:57] Amanda Vizedom: @MarkusKroetzsch - the above comment/question was a reaction to your slide 2; I     (1N2AAS)

see on slide 6 that you begin to address this.     (1N2AAT)

[9:59] Pascal Hitzler: @AmandaVizedom: I'm not Markus, of course, but let me add my own perspective     (1N2AAU)

here: The paradigms recently seem to be converging, and in particular work done and initiated by     (1N2AAV)

Markus is central for this convergence. Perhaps a good starting point for looking into this is     (1N2AAW)

Markus Kroetzsch, Frederick Maier, Adila Alfa Krisnadhi, Pascal Hitzler, A Better Uncle For OWL -     (1N2AAX)

Nominal Schemas for Integrating Rules and Ontologies. In: S. Sadagopan, Krithi Ramamritham, Arun     (1N2AAY)

Kumar, M.P. Ravindra, Elisa Bertino, Ravi Kumar (eds.), WWW '11 20th International World Wide Web     (1N2AAZ)

Conference, Hyderabad, India, March / April 2011. ACM, New York, 2011, pp. 645-654.     (1N2AAAA)

[9:58] Michael Grüninger: Speaking controversially, there are many people who equate ontologies with     (1N2AAAC)

logical theories which are definable in OWL or RDF, and for such people, anything expressed with     (1N2AAAD)

rules is something that is outside any ontology. Hopefully, this Ontolog mini-series will dispel     (1N2AAAE)

this misconception.     (1N2AAAF)

[9:59] Simon Spero: If you look at some of the gene and phenotype ontologies, the force fitting into     (1N2AAAG)

DL for reactions, enzymes etc is especially noticeable     (1N2AAAH)

[10:00] Pascal Hitzler: @MichaelGruninger: I very much agree :)     (1N2AAAI)

[10:00] Amanda Vizedom: +1 for dispelling that. I'm aware of it, and am happy to acknowledge that as     (1N2AAAJ)

a choice people can make in the type of ontology they work with, but it is frustrating to see people     (1N2AAAK)

re-define ontology so as to exclude much of its historical and continuing development and application!     (1N2AAAL)

[10:18] Markus Kroetzsch: @MichaelGruninger: "many people [...] equate ontologies with logical     (1N2AAAM)

theories which are definable in OWL" I have not witnessed this a lot. At least the Description Logic     (1N2AAAN)

community is quite open to this (the work I mentioned was presented at DL workshop even). I would     (1N2AAAO)

not call Existential Rules a type of Description Logic, but calling them an ontology language seems fine.     (1N2AAAP)

[10:19] Enrico Franconi: @Markus: +1     (1N2AAAQ)

[10:01] Simon Spero: @michaelgruninger: that's not real world controversial, let alone Toronto :-)     (1N2AAAR)

[10:00] anonymous morphed into Aida Gandara     (1N2AAAS)

[10:01] anonymous morphed into Naicong Li     (1N2AAAT)

[10:02] Pascal Hitzler: I freely admit that I'm usually using rules when modeling OWL. Or more     (1N2AAAU)

precisely, I start with writing rules (which is easier for my brain), or some hybrid (informal)     (1N2AAAV)

rules/DL notation and then convert them to OWL.     (1N2AAAW)

[10:07] Amanda Vizedom: Pascal, when have worked in OWL, I do the same thing. Having academic logic     (1N2AAAX)

training, then beginning my work in applied ontology with 6 years working in CycL, I also find it     (1N2AAAY)

much easier to my brain to come up with rules. Translation follows, though sometimes it is not     (1N2AAAZ)

feasible in a given OWL/ application infrastructure and some of the knowledge available for semantic     (1N2AAAAA)

capture simply gets left out. Sometimes that's OK. Sometimes it entails reuse problems down the     (1N2AAAAB)

[10:02] HectorPerezUrbina: @AmandaVizedom, I found this paper particularly helpful:     (1N2AAAAD)

[10:04] Amanda Vizedom: Thanks, @Hector. Reconciliation, hmmm. :-)     (1N2AAAAF)

[10:06] Pascal Hitzler: Boris' paper is a landmark. There's some work tightening the integration     (1N2AAAAG)

which follows up on this, e.g. Matthias Knorr, Pascal Hitzler, Frederick Maier, Reconciling OWL and     (1N2AAAAH)

Non-monotonic Rules for the Semantic Web. In: De Raedt, L., Bessiere, C., Dubois, D., Doherty, P.,     (1N2AAAAI)

Frasconi, P., Heintz, F., Lucas, P. (eds.), ECAI 2012, 20th European Conference on Artificial     (1N2AAAAJ)

Intelligence, 27-31 August 2012, Montpellier, France. Frontiers in Artificial Intelligence and     (1N2AAAAK)

Applications, Vol. 242, IOS Press, Amsterdam, 2012, pp. 474-479.     (1N2AAAAL)

[10:01] Enrico Franconi: (A technical question: is VNC up? I can't connect)     (1N2AAAAN)

[10:02] Pascal Hitzler: @EnricoFranconi: I'm using VNC, so it's working for me     (1N2AAAAO)

[10:02] Peter P. Yim: @EnricoFranconi - don't worry, some people cannot connect properly to the vnc     (1N2AAAAP)

server ... just use you local slides, and remember to prompt slide advances and the slide number verbally     (1N2AAAAQ)

[10:06] Michael Grüninger: @MarkusKroetzsch: Are the rules used in the chemistry example on slide 12     (1N2AAAAS)

available online?     (1N2AAAAT)

[10:08] Markus Kroetzsch is trying to catch up with the chat now ...     (1N2AAAAV)

[10:09] Amanda Vizedom: @MarkusKroetzsch: By no tools / libraries, did you mean no *open source*     (1N2AAAAW)

tools libraries? I agree with the latter but not the former, obviously.     (1N2AAAAX)

[10:16] Markus Kroetzsch: @AmandaVizedom: I mainly meant that the rule-related tools/libraries world     (1N2AAAAY)

is quite fragmented. There is RIF, which I am not aware of libraries for. Almost all reasoners that     (1N2AAAAZ)

can handle Datalog and its extensions are from the ASP and LP world, using some (more or less     (1N2AAAAAA)

uniform) Prolog-style syntax, which I don't know how to specify datatypes and URI-based identifiers     (1N2AAAAAB)

in. The ontology world has SWRL but the tools that support this are AFAIK not happy with some 80k     (1N2AAAAAC)

rules as in my example. And LP tools do not support SWRL or RIF syntax.     (1N2AAAAAD)

[10:10] Pascal Hitzler: @AmandaVizedom: For me, the cases where we're *not* able to make the     (1N2AAAAAE)

transformation into OWL, are particularly interesting (as a researcher). They point towards     (1N2AAAAAF)

limitations of the OWL standard which are worthwhile to work on in attempts to overcome them :)     (1N2AAAAAG)

[10:11] Amanda Vizedom: @Pascal: I agree! I also think that research to identify the patterns of     (1N2AAAAAH)

these cases would be both interesting and valuable. The value I'm thinking of is particularly in the     (1N2AAAAAI)

area of guidance for choosing ontology types for particular uses.     (1N2AAAAAJ)

[10:10] anonymous morphed into Ryan Hohimer     (1N2AAAAAK)

[10:11] Enrico Franconi: I guess that the real challenge with rules as an ontology language is their     (1N2AAAAAL)

integration with more classical FOL-based ontology languages (such as OWL and stuff). Obvious     (1N2AAAAAM)

mismatches, as already noticed, are closed vs open world assumption, standard/unique name     (1N2AAAAAN)

assumption, active domains, etc.     (1N2AAAAAO)

[10:12] Ali Hashemi: Is this non-standard language. Or OWL-centric thinking? Axioms don't have     (1N2AAAAAP)

variables? I guess it's a very specific notion of axiom?     (1N2AAAAAQ)

[10:13] Michael Grüninger: @EnricoFranconi: You comment implies that you are assuming that rules are     (1N2AAAAAR)

necessarily nonmonotonic. One can also consider rules to simply be a syntactic restriction with a     (1N2AAAAAS)

monotonic semantics.     (1N2AAAAAT)

[10:13] Amanda Vizedom: @EnricoFranconi: Do you mean rules in a particular syntax? I ask because     (1N2AAAAAU)

classical and current ontology languages that are based on FOL or HOL of course have rule     (1N2AAAAAV)

representation (and use) as an integral part of them.     (1N2AAAAAW)

[10:14] Harold Boley: RE Slide 4: SWRL can also be serialized in RuleML/XML (rather than in RDF/XML).     (1N2AAAAAX)

[11:02] HectorPerezUrbina: @Harold, I've just seen your comments; thank you very much for your pointers.     (1N2AAAAAZ)

[10:15] Enrico Franconi: @Hector: why are you ignoring the W3C standard RIF syntax?     (1N2AAAAAAA)

[10:17] Amanda Vizedom: Regarding Open / Closed World -- it's also worth noting that it is not     (1N2AAAAAAC)

necessary that a *language* make an open/closed world commitment. IMHO, this is more properly     (1N2AAAAAAD)

something that characterizes reasoning -- an inference parameter. And it can be treated that way, by     (1N2AAAAAAE)

supporting explicit declaration: either in an ontology module, stating that it should be interpreted     (1N2AAAAAAF)

with OW or CW, or in an application or particular query.     (1N2AAAAAAG)

[10:18] HassanAitKaci: Good point Amanda. But which would the default be?     (1N2AAAAAAH)

[10:18] Enrico Franconi: @AmandaVizedom: Open / Closed World assumption regards the semantics of the     (1N2AAAAAAI)

data! The different languages (and their reasoners) that operate on the data have to respect their     (1N2AAAAAAJ)

meaning, and adopt the right assumption.     (1N2AAAAAAK)

[10:24] Amanda Vizedom: @EnricoFranconi: Some models made such assumptions and some don't. I     (1N2AAAAAAL)

absolutely agree that when they do, it needs to be explicit. That is why I say that it is important     (1N2AAAAAAM)

to be able to make explicit that a particular ontology/module (or even rule) has one or the other     (1N2AAAAAAN)

assumption as part of its semantics.     (1N2AAAAAAO)

[10:27] Enrico Franconi: I insist that CWA/OWA (Closed World Assumption / Open World Assumption) is a     (1N2AAAAAAP)

property of your data: either we know all of your children or only the ones I'm certain about. I can     (1N2AAAAAAQ)

not use an OWA reasoner on top of some data which states complete knowledge about children, say. Or     (1N2AAAAAAR)

I can not use a CWA reasoner if the data I have is incomplete. We had an extensive discussion on     (1N2AAAAAAS)

this at last year ISWC: Peter F. Patel-Schneider and Enrico Franconi. Ontology constraints in     (1N2AAAAAAT)

incomplete and complete data. In ISWC 2012 - 11th International Semantic Web Conference, volume 7649     (1N2AAAAAAU)

of Lecture Notes in Computer Science, pages 444459. Springer-Verlag, 2012.     (1N2AAAAAAV)

[10:20] Amanda Vizedom: @MarkusKroetzsch: Very true in the OSS and for standard languages. This is     (1N2AAAAAAW)

IMHO a critical problem. The tools, and considerable integrated support, exist in proprietary     (1N2AAAAAAX)

software and languages. We need them in the open standard-based world, too.     (1N2AAAAAAY)

[10:21] Conrad Bock: Maybe tools could let users know when they go outside profiles as they enter rules.     (1N2AAAAAAZ)

[10:22] Markus Kroetzsch: @AmandaVizedom: Are there specific closed tools and standards you are     (1N2AAAAAAAA)

thinking about? I recall how the RIF Working Group decided to ignore the ISO Prolog standard because     (1N2AAAAAAAB)

most group members had no access to this closed standard. But many Prolog tools are free.     (1N2AAAAAAAC)

[10:27] Amanda Vizedom: @Markus: CycL / the Cyc system have the most comprehensive support for all of     (1N2AAAAAAAD)

this that I have worked with. However, I've also worked with a number of organizations that have     (1N2AAAAAAAE)

developed their own KR language and tools, for internal use, that have some subset of these thing     (1N2AAAAAAAF)

that fit what they need for their application type(s) and typical domain problem.     (1N2AAAAAAAG)

[10:22] Simon Spero: @HassanAitKaci: CWA/OWA in Cyc is configurable on a per predicate basis, with     (1N2AAAAAAAH)

NAF (negation as failure) settable on per query basis with default of off     (1N2AAAAAAAI)

[10:23] HassanAitKaci: @SimonSpero - good for Cyc then ...     (1N2AAAAAAAJ)

[10:31] Harold Boley: The relational SWRL Built-Ins were complemented by the functional RIF     (1N2AAAAAAAL)

[10:32] Amanda Vizedom: @EnricoFranconi: In domains in which the data are almost always incomplete,     (1N2AAAAAAAO)

it is quite common to want to reason with that unknownness explicit in some contexts and for other     (1N2AAAAAAAP)

purposes to reason as if (some of) the data were complete, to explore hypotheses and possibilities.     (1N2AAAAAAAQ)

Again, why do you think that when it is a property of the data, it cannot be simply stated and     (1N2AAAAAAAR)

reasoned on appropriately? Why make that a fixed feature of a language or reasoner instead?     (1N2AAAAAAAS)

[10:35] HectorPerezUrbina: @Amanda, suppose you want to enforce that every instance of the class     (1N2AAAAAAAT)

Employee in your ontology has to have a SSN; under OWA, we couldn't enforce this. We'd have to     (1N2AAAAAAAU)

resort to CWA to be able to find this kind of violation. However, on the other hand, I would still     (1N2AAAAAAAV)

want to infer (under open world) that all employees are people.     (1N2AAAAAAAW)

[10:35] HectorPerezUrbina: As Enrico says, whether you want to use OWA or CWA, depends on the nature     (1N2AAAAAAAX)

of your data, regardless of the specific formalism you're using to model it     (1N2AAAAAAAY)

[10:36] Enrico Franconi: Well, what I'm saying is that you may want to consider your *data* open or     (1N2AAAAAAAZ)

closed - the choice is up to you . BUT you should never use different assumptions at the same time     (1N2AAAAAAAAA)

over the same set of data.     (1N2AAAAAAAAB)

[10:36] Markus Kroetzsch: @HectorPerezUrbina: ... and on the context (the same data might be OWA when     (1N2AAAAAAAAC)

viewed as an ontology and CWA when viewed as a list of axioms that I want to search through)     (1N2AAAAAAAAD)

[10:36] Markus Kroetzsch: @AmandaVizedom: Can you recommend any references on the CWA/OWA combination     (1N2AAAAAAAAE)

that you refer to?     (1N2AAAAAAAAF)

[10:37] HectorPerezUrbina: oh, I disagree. Using the previous example, you could first use OWA to     (1N2AAAAAAAAG)

infer that Hector is an Employee, and then CWA to determine that, since Hector doesn't have a SSN,     (1N2AAAAAAAAH)

the ontology is invalid.     (1N2AAAAAAAAI)

[10:37] Amanda Vizedom: @Hector, of the data *and* of the reasoning you want to do, no? Again, I'm     (1N2AAAAAAAAJ)

not arguing that the assumptions should be ignored; I'm arguing that they should be explicitly     (1N2AAAAAAAAK)

statable and the reasoners should be able to understand that and use it.     (1N2AAAAAAAAL)

[10:37] Simon Spero: @EnricoFranconi: agree if it's just assumptions that you're talking about - if     (1N2AAAAAAAAM)

an epistemic axiom is asserted, then it stops being an assumption     (1N2AAAAAAAAN)

[10:38] Simon Spero: Driveby coverage in 12.8 of     (1N2AAAAAAAAO)

[10:39] Amanda Vizedom: @Hector: Your constraint-violation case is an excellent example of a common     (1N2AAAAAAAAQ)

use case that comes up with OWL ontologies. when you are looking for (potential) violations, you may     (1N2AAAAAAAAR)

want to temporarily use CWA, rather than use OWA and infer missing statements.     (1N2AAAAAAAAS)

[10:40] HectorPerezUrbina: @Amanda, certainly. In Stardog, for example, we allow to validate     (1N2AAAAAAAAT)

integrity constraints (under CWA) and usual OWL inference (under OWA), but it is necessary for the     (1N2AAAAAAAAU)

user to clearly specify which things are to be taken under which semantics.     (1N2AAAAAAAAV)

[10:39] Amanda Vizedom: Okay, I need to stop typing and pay attention to Hassan for a bit!     (1N2AAAAAAAAW)

[10:40] Pascal Hitzler: Regarding the open/closed world discussion: What is really needed are     (1N2AAAAAAAAX)

languages which combine open world and closed world features, in such a way that you can decide     (1N2AAAAAAAAY)

what's supposed to be closed and what's supposed to be open. The keyword sometimes used is "local     (1N2AAAAAAAAZ)

closed world". There's quite a bit of work under way how to do this (see e.g. the paper by Motik     (1N2AAAAAAAAAA)

referenced above), but the quest is not yet conclusive, in particular in practical terms.     (1N2AAAAAAAAAB)

[10:40] Amanda Vizedom: @PascalHitzler: Exactly!     (1N2AAAAAAAAAC)

[10:41] Amanda Vizedom: @Hector: Excellent. :-)     (1N2AAAAAAAAAD)

[10:41] HectorPerezUrbina: @Amanda, you might find this interesting: http://docs.stardog.com/icv/     (1N2AAAAAAAAAE)

[10:42] Amanda Vizedom: Thanks Hector, I'll check it out.     (1N2AAAAAAAAAF)

[10:45] Enrico Franconi: Again, in my paper I argue why epistemic axioms can be very misleading. In     (1N2AAAAAAAAAG)

your simple example with SSN, this approach may work. But since you expect two different beahviours     (1N2AAAAAAAAAH)

from your data stemming form conflicting assumptions, there will always be counterexamples where     (1N2AAAAAAAAAI)

your get unexpected inferences.     (1N2AAAAAAAAAJ)

[10:46] Enrico Franconi: @Hector: really look at our paper on why you shouldn't validate integrity     (1N2AAAAAAAAAK)

constraints with a different semantics from the underlying ontology language.     (1N2AAAAAAAAAL)

[10:46] HectorPerezUrbina: @Enrico, yes, it helps that we consider relatively unexpressive logics     (1N2AAAAAAAAAM)

(i.e., OWL profiles)     (1N2AAAAAAAAAN)

[10:47] Markus Kroetzsch: +1 to epistemic axioms being confusing     (1N2AAAAAAAAAO)

[10:47] Enrico Franconi: Mmhh, I guess I can build some nasty counterexamples with dl-lite as well,     (1N2AAAAAAAAAQ)

but I have to think about it.     (1N2AAAAAAAAAR)

[10:48] HectorPerezUrbina: @Enrico, will do. I would say, however, that we are yet to see these     (1N2AAAAAAAAAS)

nasty examples in practice.     (1N2AAAAAAAAAT)

[10:48] Amanda Vizedom: @EnricoFranconi: Do you have a link to your paper? I would like to understand     (1N2AAAAAAAAAV)

why you think that support for making assumptions explicit is inferior to being restricted to one of     (1N2AAAAAAAAAW)

the assumptions. ... ref. below - [11:13] Enrico Franconi: My paper on "Ontology constraints in     (1N2AAAAAAAAAX)

[10:51] Amanda Vizedom: Epistemic axioms may be confusing, but IMHO the better way of addressing this     (1N2AAAAAAAAAZ)

is to make clear and explicit which axioms are epistemic and what they apply to (even if as     (1N2AAAAAAAAAAA)

metadata). Otherwise, users and modelers tend to *make* epistemic assumptions anyway, without making     (1N2AAAAAAAAAAB)

the explicit, and sometimes slide between epistemic assumptions within the same model or set of     (1N2AAAAAAAAAAC)

models, without being able to indicate this. This causes problems for model accuracy, usability,     (1N2AAAAAAAAAAD)

evaluation, quality control / truth maintenance (as the model evolves), and reuse.     (1N2AAAAAAAAAAE)

[10:51] Simon Spero: Hector: If every employee has a known ssn and hector does not have a known ssn     (1N2AAAAAAAAAAG)

then hector is not an employee     (1N2AAAAAAAAAAH)

[10:52] HectorPerezUrbina: @Simon, that's under OWA     (1N2AAAAAAAAAAI)

{{{ [10:54] HectorPerezUrbina: Consider the ontology O = {1. Employee subClassOf hasSSN some SSN, 2. Hector a Employee} . Under OWA, we would infer the existence of an anonymous individual, instance of SSN, related to Hector via hasSSN. }}}     (1N2AAAAAAAAAAJ)

[10:54] HectorPerezUrbina: However, sometimes, we don't want this behavior; instead, we want the     (1N2AAAAAAAAAAK)

reasoner to let us know that our data is invalid (because every employee must have a SSN)     (1N2AAAAAAAAAAL)

[10:55] HectorPerezUrbina: We can accomplish this by interpreting axiom 2 in O under CWA.     (1N2AAAAAAAAAAM)

[10:58] Amanda Vizedom: @Hector: yes, and more: we may want to validate/ declare that invalid. Or, we     (1N2AAAAAAAAAAO)

may want to identify gaps in our knowledge. Or, in more sophisticated reasoning, analyze our model     (1N2AAAAAAAAAAP)

to identify *patterns* of missing information. The known unknowns can be very important in some applications!     (1N2AAAAAAAAAAQ)

[11:05] HectorPerezUrbina: @Amanda, yes. In practice, people produce some RDF out of an ETL process,     (1N2AAAAAAAAAAR)

and they need to check whether certain integrity constraints (a la DB) hold.     (1N2AAAAAAAAAAS)

[10:59] Simon Spero: Hector: But you leave out the axiom hasSSN(X) -> K[hasSSN(X)]     (1N2AAAAAAAAAAT)

[11:00] Mike Bennett: OWA and CWA: it seems to me that the use case of determining whether some data     (1N2AAAAAAAAAAU)

is valid (per @Hector above), and the use case whereby we want to reason over assertions about real     (1N2AAAAAAAAAAV)

things in the world (not data) are two very different requirements. My hunch would be that to talk     (1N2AAAAAAAAAAW)

about data rather than actual things, one must need the CWA?     (1N2AAAAAAAAAAX)

[11:03] HectorPerezUrbina: @Simon, yes, we want to stay within OWL and SWRL     (1N2AAAAAAAAAAY)

[11:03] HassanAitKaci: @EnricoFranconi: This is a CWA construction! :-)     (1N2AAAAAAAAAAZ)

[10:57] Alan Rector: [ref. slide#5] Shouldn't be that some ice cream is food, rather than that all     (1N2AAAAAAAAAAAA)

ice cream is food/ice cream is subset of food?     (1N2AAAAAAAAAAAB)

[10:58] Pascal Hitzler: @Alan, I believe in the example, ice cream is an individual, while food is a class?     (1N2AAAAAAAAAAAC)

[11:00] Francesca Quattri: @EnricoFranconi: Beth Levin's work could provide a great deal of     (1N2AAAAAAAAAAAD)

inspiration for defining what is (or should be defined) subclass of what     (1N2AAAAAAAAAAAE)

[11:00] Mike Bennett: Apologies, have to leave now.     (1N2AAAAAAAAAAAF)

[11:04] anonymous morphed into Nancy Wiegand     (1N2AAAAAAAAAAAG)

[11:07] Alan Rector: Apologies. I have to leave now     (1N2AAAAAAAAAAAH)

[11:07] Peter P. Yim: == Q&A and open discussion now     (1N2AAAAAAAAAAAI)

[11:09] anonymous morphed into Dennis Wisnosky     (1N2AAAAAAAAAAAO)

[11:10] anonymous1 morphed into lakhdar     (1N2AAAAAAAAAAAP)

[11:11] Amanda Vizedom: @Hector, this is one type of case. But I am also thinking of another case in     (1N2AAAAAAAAAAAQ)

which a user is using an ontological KB for situation awareness, risk analysis, or similar. In these     (1N2AAAAAAAAAAAR)

cases, while much of the use of the model involves reasoning about the modeled things (or digesting     (1N2AAAAAAAAAAAS)

views of a situation that are based on model), it can be an essential part of this use to also     (1N2AAAAAAAAAAAT)

(manually or automatically in side processes) identify the known unknowns and spot patterns in them.     (1N2AAAAAAAAAAAU)

This can be critical.     (1N2AAAAAAAAAAAV)

[11:12] HectorPerezUrbina: @Amanda, yes, this sounds quite interesting/important.     (1N2AAAAAAAAAAAW)

[11:13] Enrico Franconi: My paper on "Ontology constraints in incomplete and complete data" can be     (1N2AAAAAAAAAAAX)

[11:13] HectorPerezUrbina: @Enrico, thank you     (1N2AAAAAAAAAAAZ)

[11:13] Amanda Vizedom: @EnricoFranconi: Thanks, I will read it.     (1N2AAAAAAAAAAAAA)

[11:16] Enrico Franconi: Just do (:yourLocalProperty rdfs:subProperty :globalProperty)!     (1N2AAAAAAAAAAAAB)

[11:20] Enrico Franconi: [in response to PatrickMaroney's question on where details of     (1N2AAAAAAAAAAAAD)

EnricoFranconi's work (the evaluations, in particular) can be accessed] Official link:     (1N2AAAAAAAAAAAAE)

[11:16] Peter P. Yim: @EnricoFranconi - (recapping my verbal comment) it would be great to further     (1N2AAAAAAAAAAAAH)

expose your work to Pat Hayes, and convince him and those who are working on the next iteration of     (1N2AAAAAAAAAAAAI)

RDF/RDFS, so your work can be taken into consideration in that W3C standard ... [I will try to     (1N2AAAAAAAAAAAAJ)

forward this to Pat Hayes, copying you too, after the session]     (1N2AAAAAAAAAAAAK)

[11:17] Enrico Franconi: @PeterYim: thanks     (1N2AAAAAAAAAAAAL)

[11:22] Simon Spero: RDF Semantics 1.1 just went to Candidate Rec status on Nov 5th, ( last step     (1N2AAAAAAAAAAAAM)

[11:22] Enrico Franconi: "Work as expected" means complete & terminating in the fragment of the use case.     (1N2AAAAAAAAAAAAO)

[11:25] Markus Kroetzsch: Sorry, I got distracted by my phone call. To clarify: in the context of     (1N2AAAAAAAAAAAAQ)

ontology "undecidable" = "not sufficiently studied to understand when it will work" (it's different     (1N2AAAAAAAAAAAAR)

for programming and problem solving languages, where undecidability is necessary to express     (1N2AAAAAAAAAAAAS)

arbitrarily complex computations/problems.)     (1N2AAAAAAAAAAAAT)

[11:26] Harold Boley: Regarding decidable <> efficient etc., average-case complexity should be     (1N2AAAAAAAAAAAAU)

considered, rather than (only) worst-case complexity.     (1N2AAAAAAAAAAAAV)

[11:27] Enrico Franconi: @Harold: yes, but sill in the context of complete and terminating fragments.     (1N2AAAAAAAAAAAAY)

[11:27] HectorPerezUrbina: also, so-called data complexity is quite important     (1N2AAAAAAAAAAAAZ)

[11:26] Peter P. Yim: Join us again, in two weeks (Thu 2013-12-05) for the Ontology Summit 2014 Pre-Launch     (1N2AAAAAAAAAAAAAC)

Community Session, when we will collaboratively work up a program for the next OntologySummit.     (1N2AAAAAAAAAAAAAD)

[11:26] Peter P. Yim: That will be followed (on Dec-12) by session-07 of the OntologyBasedStandards     (1N2AAAAAAAAAAAAAE)

miniseries - "How ontologies can help with the formal specification of the natural language     (1N2AAAAAAAAAAAAAF)

[11:26] Peter P. Yim: The next event (session-04) for this RulesReasoningLP mini-series will then come up     (1N2AAAAAAAAAAAAAH)

on Dec-19 - "Guide to Reasoning Applications Development and Cases" - Co-chairs: Henson Graves & Ken Baclawski     (1N2AAAAAAAAAAAAAI)

[11:28] Leo Obrst: Thanks, All! Very interesting session.     (1N2AAAAAAAAAAAAAN)

[11:28] Peter P. Yim: -- session ended: 11:28am PST --     (1N2AAAAAAAAAAAAAQ)

-- end of in-session chat-transcript --     (1N2AAAAAAAAAAAAAR)

Additional Resources     (1O)


For the record ...     (1O4)

How To Join (while the session is in progress)     (1P)


This page has been migrated from the OntologWiki - Click here for original page     (1P4)