From f.rabe at jacobs-university.de Fri Feb 6 15:02:49 2015 From: f.rabe at jacobs-university.de (Florian Rabe) Date: Fri, 6 Feb 2015 15:02:49 +0100 Subject: [Hets-devel] Trade-off between theory and model morphisms? Message-ID: <18983A3FED046183EEFEEFA4@[10.50.250.235]> Dear all, It seems not every LF-style theory morphism induces a model reduction function. Maybe Till or others have known this for a while, but it only occurred to me now. Details are below. I'm interested in any corrections, comments, or literature pointers. Ciao, Florian 1) Using LF, we've so far defined L-theory morphisms t: S -> T as LF-morphisms L,S -> L,T that fix L. For L = FOL, this means that t must map every n-ary S-predicate symbol p to a T-formula F with n free variables. 2) For FOL-style logics, a model morphism m: M -> N - maps between the universes - commutes with function symbols - preserves predicate symbols [1], i.e., p^M(u) = 1 => p^N(m(u)) = 1 3) Now those two definitions do not fit together: Not every theory morphism t can reduce every T-model morphism m. The problem occurs when t maps an S-predicate symbol p to a T-formula that is not preserved [2] by T-model morphisms. [3] That does not affect truth or theoremhood but does means, e.g., that we can have isomorphic theories with non-isomorphic model categories. Our previous results should not be affected because we had never figured anyway out how to define model morphisms using LF. But it means the model morphism problem is even harder than I thought. Notes: [1] Requiring <=> instead of => would be more natural from the LF-perspective because it would treat function and predicate symbols uniformly. But the problem would still persist. [2] Conjunctions, disjunctions, and existentials of atomic formulas are preserved. [3] [3] Not affected are CASL-style derived morphisms S -> T' where we extend T to T' with a defined predicate symbol p'(x)<=>F(x) and then map p to p'. [4] [4] Thus, adding a defined predicate symbol can change the model category. T -> T' is a conservative extension with unique model extension. Yet, not every T-model morphism is a T'-model morphism. From mossakow at iws.cs.uni-magdeburg.de Sat Feb 7 09:39:29 2015 From: mossakow at iws.cs.uni-magdeburg.de (Till Mossakowski) Date: Sat, 07 Feb 2015 09:39:29 +0100 Subject: [Hets-devel] [LATIN] Trade-off between theory and model morphisms? In-Reply-To: <18983A3FED046183EEFEEFA4@[10.50.250.235]> References: <18983A3FED046183EEFEEFA4@[10.50.250.235]> Message-ID: <54D5CF41.10602@iws.cs.uni-magdeburg.de> Hi Florian, [I am adding Ulf and Tom into cc because this problem has been ignored in our joint paper.] Am 06.02.2015 15:02, schrieb Florian Rabe: > Dear all, > > It seems not every LF-style theory morphism induces a model reduction > function. > Maybe Till or others have known this for a while, but it only occurred > to me now. > > Details are below. > I'm interested in any corrections, comments, or literature pointers. > > Ciao, > Florian > > 1) Using LF, we've so far defined L-theory morphisms t: S -> T as > LF-morphisms L,S -> L,T that fix L. > > For L = FOL, this means that t must map every > n-ary S-predicate symbol p to a T-formula F with n free variables. > > 2) For FOL-style logics, a model morphism m: M -> N > - maps between the universes > - commutes with function symbols > - preserves predicate symbols [1], i.e., > p^M(u) = 1 => p^N(m(u)) = 1 > > 3) Now those two definitions do not fit together: Not every theory > morphism t can reduce every T-model morphism m. > > The problem occurs when t maps an S-predicate symbol p to a T-formula > that is not preserved [2] by T-model morphisms. [3] > > That does not affect truth or theoremhood but does means, e.g., that > we can have isomorphic theories with non-isomorphic model categories. this is of course bad. It implies that MOD (taking theories to model categories) is not functorial. > > Our previous results should not be affected because we had never > figured anyway out how to define model morphisms using LF. > But it means the model morphism problem is even harder than I thought. > > Notes: > > [1] Requiring <=> instead of => would be more natural from the > LF-perspective because it would treat function and predicate symbols > uniformly. > But the problem would still persist. However, => is more widely used for FOL. I am just reviewing a PhD thesis about spatial reasoning and constraint satisfaction where this appears again. > > [2] Conjunctions, disjunctions, and existentials of atomic formulas > are preserved. [3] Actually, Razvan Diaconescu calls these formulas "atomic", see his book "Institution-independent model theory". > > [3] Not affected are CASL-style derived morphisms S -> T' where we > extend T to T' with a defined predicate symbol p'(x)<=>F(x) and then > map p to p'. [4] attached a recent paper about derived morphisms. Feedback is welcome. Indeed, in this paper we have entirely ignored model morphisms. But we should run into the same problem. > > [4] Thus, adding a defined predicate symbol can change the model > category. > T -> T' is a conservative extension with unique model extension. > Yet, not every T-model morphism is a T'-model morphism. Indeed, this only holds if F(x) is an "atomic" formula in the above sense. Hence, One solution is to restrict formulas in derived morphisms to such "atomic" formulas. Another one is to live with the fact that derived signatures only induce reducts between the discretised model categories. Both are a bit unsatisfactory. Best, Till > _______________________________________________ > project-LATIN mailing list > project-LATIN at lists.jacobs-university.de > http://lists.jacobs-university.de/mailman/listinfo/project-latin -------------- next part -------------- A non-text attachment was scrubbed... Name: paper.pdf Type: application/pdf Size: 393466 bytes Desc: not available URL: From f.rabe at jacobs-university.de Sat Feb 7 17:37:20 2015 From: f.rabe at jacobs-university.de (Florian Rabe) Date: Sat, 7 Feb 2015 17:37:20 +0100 Subject: [Hets-devel] [LATIN] Trade-off between theory and model morphisms? Message-ID: <8657F0F80C4887FC2785E245@[192.168.1.50]> Hi Till, > [I am adding Ulf and Tom into cc because this problem has been ignored > in our joint paper.] Indeed, your definition (after Def. 8) defines the functor Mod(sigma) on objects but not on morphisms. I think it is not possible to define it on morphisms without restricting the original institution. Thm. 1 would have to be restricted accordingly. As you indicate, there appear to be two reasonable restrictions: - only allow model morphisms that preserve all sentences (unsatisfactory of course, but slightly more general than discrete model categories) - only allow definitional extensions that preserve model morphisms Best, Florian > Am 06.02.2015 15:02, schrieb Florian Rabe: >> Dear all, >> >> It seems not every LF-style theory morphism induces a model reduction >> function. >> Maybe Till or others have known this for a while, but it only occurred >> to me now. >> >> Details are below. >> I'm interested in any corrections, comments, or literature pointers. >> >> Ciao, >> Florian >> >> 1) Using LF, we've so far defined L-theory morphisms t: S -> T as >> LF-morphisms L,S -> L,T that fix L. >> >> For L = FOL, this means that t must map every >> n-ary S-predicate symbol p to a T-formula F with n free variables. >> >> 2) For FOL-style logics, a model morphism m: M -> N >> - maps between the universes >> - commutes with function symbols >> - preserves predicate symbols [1], i.e., >> p^M(u) = 1 => p^N(m(u)) = 1 >> >> 3) Now those two definitions do not fit together: Not every theory >> morphism t can reduce every T-model morphism m. >> >> The problem occurs when t maps an S-predicate symbol p to a T-formula >> that is not preserved [2] by T-model morphisms. [3] >> >> That does not affect truth or theoremhood but does means, e.g., that >> we can have isomorphic theories with non-isomorphic model categories. > this is of course bad. It implies that MOD (taking theories to model > categories) is not functorial. >> >> Our previous results should not be affected because we had never >> figured anyway out how to define model morphisms using LF. >> But it means the model morphism problem is even harder than I thought. >> >> Notes: >> >> [1] Requiring <=> instead of => would be more natural from the >> LF-perspective because it would treat function and predicate symbols >> uniformly. >> But the problem would still persist. > However, => is more widely used for FOL. I am just reviewing a PhD > thesis about spatial reasoning and constraint satisfaction where this > appears again. >> >> [2] Conjunctions, disjunctions, and existentials of atomic formulas >> are preserved. [3] > Actually, Razvan Diaconescu calls these formulas "atomic", see his book > "Institution-independent model theory". >> >> [3] Not affected are CASL-style derived morphisms S -> T' where we >> extend T to T' with a defined predicate symbol p'(x)<=>F(x) and then >> map p to p'. [4] > attached a recent paper about derived morphisms. Feedback is welcome. > Indeed, in this paper we have entirely ignored model morphisms. But we > should run into the same problem. >> >> [4] Thus, adding a defined predicate symbol can change the model >> category. >> T -> T' is a conservative extension with unique model extension. >> Yet, not every T-model morphism is a T'-model morphism. > Indeed, this only holds if F(x) is an "atomic" formula in the above > sense. Hence, One solution is to restrict formulas in derived morphisms > to such "atomic" formulas. Another one is to live with the fact that > derived signatures only induce reducts between the discretised model > categories. Both are a bit unsatisfactory. > > Best, Till >> _______________________________________________ >> project-LATIN mailing list >> project-LATIN at lists.jacobs-university.de >> http://lists.jacobs-university.de/mailman/listinfo/project-latin > From tom at maibaum.org Sat Feb 7 16:01:04 2015 From: tom at maibaum.org (Tom Maibaum) Date: Sat, 7 Feb 2015 10:01:04 -0500 Subject: [Hets-devel] [LATIN] Trade-off between theory and model morphisms? In-Reply-To: <54D5CF41.10602@iws.cs.uni-magdeburg.de> References: <18983A3FED046183EEFEEFA4@[10.50.250.235]> <54D5CF41.10602@iws.cs.uni-magdeburg.de> Message-ID: Initial reaction (I have not really had time to absorb this properly): The model theoretic def of conservative extension is: every model of the source theory OR AN ELEMENTARILY ISOMORPHIC MODEL can be extended to a model of the target theory. I would not be surprised if this is what is tripping you up ... > On Feb 7, 2015, at 3:39 AM, Till Mossakowski wrote: > > Hi Florian, > > [I am adding Ulf and Tom into cc because this problem has been ignored > in our joint paper.] > > Am 06.02.2015 15:02, schrieb Florian Rabe: >> Dear all, >> >> It seems not every LF-style theory morphism induces a model reduction >> function. >> Maybe Till or others have known this for a while, but it only occurred >> to me now. >> >> Details are below. >> I'm interested in any corrections, comments, or literature pointers. >> >> Ciao, >> Florian >> >> 1) Using LF, we've so far defined L-theory morphisms t: S -> T as >> LF-morphisms L,S -> L,T that fix L. >> >> For L = FOL, this means that t must map every >> n-ary S-predicate symbol p to a T-formula F with n free variables. >> >> 2) For FOL-style logics, a model morphism m: M -> N >> - maps between the universes >> - commutes with function symbols >> - preserves predicate symbols [1], i.e., >> p^M(u) = 1 => p^N(m(u)) = 1 >> >> 3) Now those two definitions do not fit together: Not every theory >> morphism t can reduce every T-model morphism m. >> >> The problem occurs when t maps an S-predicate symbol p to a T-formula >> that is not preserved [2] by T-model morphisms. [3] >> >> That does not affect truth or theoremhood but does means, e.g., that >> we can have isomorphic theories with non-isomorphic model categories. > this is of course bad. It implies that MOD (taking theories to model > categories) is not functorial. >> >> Our previous results should not be affected because we had never >> figured anyway out how to define model morphisms using LF. >> But it means the model morphism problem is even harder than I thought. >> >> Notes: >> >> [1] Requiring <=> instead of => would be more natural from the >> LF-perspective because it would treat function and predicate symbols >> uniformly. >> But the problem would still persist. > However, => is more widely used for FOL. I am just reviewing a PhD > thesis about spatial reasoning and constraint satisfaction where this > appears again. >> >> [2] Conjunctions, disjunctions, and existentials of atomic formulas >> are preserved. [3] > Actually, Razvan Diaconescu calls these formulas "atomic", see his book > "Institution-independent model theory". >> >> [3] Not affected are CASL-style derived morphisms S -> T' where we >> extend T to T' with a defined predicate symbol p'(x)<=>F(x) and then >> map p to p'. [4] > attached a recent paper about derived morphisms. Feedback is welcome. > Indeed, in this paper we have entirely ignored model morphisms. But we > should run into the same problem. >> >> [4] Thus, adding a defined predicate symbol can change the model >> category. >> T -> T' is a conservative extension with unique model extension. >> Yet, not every T-model morphism is a T'-model morphism. > Indeed, this only holds if F(x) is an "atomic" formula in the above > sense. Hence, One solution is to restrict formulas in derived morphisms > to such "atomic" formulas. Another one is to live with the fact that > derived signatures only induce reducts between the discretised model > categories. Both are a bit unsatisfactory. > > Best, Till >> _______________________________________________ >> project-LATIN mailing list >> project-LATIN at lists.jacobs-university.de >> http://lists.jacobs-university.de/mailman/listinfo/project-latin > > From m_chaib at esi.dz Fri Feb 13 13:15:26 2015 From: m_chaib at esi.dz (Mostefa CHAIB) Date: Fri, 13 Feb 2015 13:15:26 +0100 Subject: [Hets-devel] Entry to Hets Message-ID: I am new in-hets devel. I think that to understand Hets better we have do know how is built. If you can point me to integrate a new language respecting the structure in Hets. - The first file to code and where to put it - The nommination files, which is principally that of grammar - Required Folder names - How to test the work within Hets - The other steps in brief ... Thanks -------------- next part -------------- An HTML attachment was scrubbed... URL: From mossakow at iws.cs.uni-magdeburg.de Sun Feb 15 17:07:25 2015 From: mossakow at iws.cs.uni-magdeburg.de (Till Mossakowski) Date: Sun, 15 Feb 2015 17:07:25 +0100 Subject: [Hets-devel] Entry to Hets In-Reply-To: References: Message-ID: <54E0C43D.7000203@iws.cs.uni-magdeburg.de> Dear Mostefa, please have a look at https://github.com/spechub/Hets and http://www.informatik.uni-bremen.de/agbkb/forschung/formal_methods/CoFI/hets/src-distribution/versions/Hets/docs/ Best, Till Am 13.02.2015 um 13:15 schrieb Mostefa CHAIB: > I am new in-hets devel. > I think that to understand Hets better we have do know how is built. > If you can point me to integrate a new language respecting the > structure in Hets. > - The first file to code and where to put it > - The nommination files, which is principally that of grammar > - Required Folder names > - How to test the work within Hets > - The other steps in brief > ... > Thanks > > > _______________________________________________ > Hets-devel mailing list > Hets-devel at informatik.uni-bremen.de > https://mailman.informatik.uni-bremen.de/mailman/listinfo/hets-devel -------------- next part -------------- An HTML attachment was scrubbed... URL: From fneuhaus at web.de Thu Feb 26 18:19:16 2015 From: fneuhaus at web.de (Fabian Neuhaus) Date: Thu, 26 Feb 2015 12:19:16 -0500 Subject: [Hets-devel] retrieving signature Message-ID: Hi all, Is there a way to retrieve a signature of an ontology in a machine readable way, preferably in Hets Interactive mode? (Or alternative, is there a command to check whether a symbol is in the signature of a given node?) If not, can we use "show-node-info-current? and use the part which reads new symbols: { ? } Or is this something else than the signature? Best Fabian From mossakow at iws.cs.uni-magdeburg.de Thu Feb 26 19:05:34 2015 From: mossakow at iws.cs.uni-magdeburg.de (Till Mossakowski) Date: Thu, 26 Feb 2015 19:05:34 +0100 Subject: [Hets-devel] retrieving signature In-Reply-To: References: Message-ID: <54EF606E.8010003@iws.cs.uni-magdeburg.de> Hi Fabian, indeed, for OWL,, the set of symbols is the signature (for CASL, it is not, due to subsort relations). However, the new symbols are only those locally introduced in the node. That is, the symbols imported from other nodes (theories) are not shown. If you need also the imported symbols, this should be easy to implement. (Of course, you can also use show-computed-theory and ignore the axioms.) Best, Till Am 26.02.2015 18:19, schrieb Fabian Neuhaus: > Hi all, > > Is there a way to retrieve a signature of an ontology in a machine readable way, preferably in Hets Interactive mode? (Or alternative, is there a command to check whether a symbol is in the signature of a given node?) > > If not, can we use "show-node-info-current? and use the part which reads > new symbols: > { > ? } > Or is this something else than the signature? > > Best > Fabian > _______________________________________________ > Hets-devel mailing list > Hets-devel at informatik.uni-bremen.de > https://mailman.informatik.uni-bremen.de/mailman/listinfo/hets-devel > From c.maeder at jacobs-university.de Thu Feb 26 19:30:31 2015 From: c.maeder at jacobs-university.de (Christian Maeder) Date: Thu, 26 Feb 2015 19:30:31 +0100 Subject: [Hets-devel] retrieving signature In-Reply-To: <54EF606E.8010003@iws.cs.uni-magdeburg.de> References: <54EF606E.8010003@iws.cs.uni-magdeburg.de> Message-ID: <54EF6647.9050509@jacobs-university.de> Hi, the xml output "hets -o xml ..", possibly extended using "--full-signatures", is supposed for machine readable output. "hets -o sig ..." prints the signature (natively), i.e. for OWL in OMN, in a separate file for each node. I believe "new symbols" are the difference of the final signature (that is closed) and a imported or extended signature. (A subsort relation is also a kind of symbol, see https://github.com/spechub/Hets/blob/master/CASL/Sign.hs "SubsortAsItemType") Checking for a symbol is not supported. (It would require to parse and check a symbol given as string in a suitable context.) Cheers Christian Am 26.02.2015 um 19:05 schrieb Till Mossakowski: > Hi Fabian, > > indeed, for OWL,, the set of symbols is the signature (for CASL, it is > not, due to subsort relations). > However, the new symbols are only those locally introduced in the node. > That is, the symbols imported from other nodes (theories) are not shown. > If you need also the imported symbols, this should be easy to implement. > (Of course, you can also use show-computed-theory and ignore the axioms.) > > Best, Till > > Am 26.02.2015 18:19, schrieb Fabian Neuhaus: >> Hi all, >> >> Is there a way to retrieve a signature of an ontology in a machine readable way, preferably in Hets Interactive mode? (Or alternative, is there a command to check whether a symbol is in the signature of a given node?) >> >> If not, can we use "show-node-info-current? and use the part which reads >> new symbols: >> { >> ? } >> Or is this something else than the signature? >> >> Best >> Fabian >> _______________________________________________ >> Hets-devel mailing list >> Hets-devel at informatik.uni-bremen.de >> https://mailman.informatik.uni-bremen.de/mailman/listinfo/hets-devel >> > > _______________________________________________ > Hets-devel mailing list > Hets-devel at informatik.uni-bremen.de > https://mailman.informatik.uni-bremen.de/mailman/listinfo/hets-devel > From m_chaib at esi.dz Fri Feb 13 07:49:25 2015 From: m_chaib at esi.dz (Mostefa CHAIB) Date: Fri, 13 Feb 2015 06:49:25 -0000 Subject: [Hets-devel] New developer Message-ID: I am new in-hets devel. I think that to understand Hets better we have do know how is built. If you can point me to integrate a new language respecting the structure in Hets. - The first file to code and where to put it - The nommination files, which is principally that of grammar - Required Folder names - How to test the work within Hets - The other steps in brief ... Thanks -------------- next part -------------- An HTML attachment was scrubbed... URL: