00:00:32 is ":-" still active in N3 whilst we're killing off bits? 00:00:49 That I have to think about hte philosophy. 00:00:53 I have to go now. 00:00:59 Isn't consensus neat? 00:01:07 :-) 00:01:09 As long as it's MY consensus :) 00:01:10 'tis a beautiful thing 00:01:17 If people discuss N3 more, please send me pointers! 00:01:28 * timbl goes 00:01:47 * bijan wonders if a description of n3 inference is in the offing... 00:01:57 perhaps this should be chumped? 00:01:58 I guess that'd be CWM inference, actually. 00:02:43 Hmm... should'a changed those "[ '#ora' >- dc:wrote-> [ >- dc:title -> "Moby Dick" ] ] ." type examples at the bottom 00:04:25 yeeech. I can't read that even though I know what it means now :) 00:04:56 [ :ora dc:creator [ dc:title "Moby Dick" ] ] . 00:05:03 Yay! 00:05:04 Doesn't wrap. 00:05:08 ooh, actually... 00:05:11 :ora dc:creator [ dc:title "Moby Dick" ] . 00:05:20 * sbp smacks self 00:06:24 ugh... in fact, it's the daml:inverseOf dc:creator: :ora [ daml:inverseOf dc:creator ] [ dc:title "Moby Dick" ] . 00:07:11 heh, "Ora was created by the thing with the title 'Moby Dick'" 00:07:54 N3 makes inverses easier: :ora is dc:creator of [ dc:title "Moby Dick" ] . 00:08:37 * bijan doesn't like that. 00:08:45 Kill is...of! 00:09:02 And has. 00:09:10 Noise, pure noise. 00:09:13 I agree with getting rid of has 00:09:25 has is a has been 00:09:43 is :x of does seem to be quite useful 00:09:54 is..of has the very undesireable property of inverting things ;) 00:10:15 why is that undesirable? sometimes, it really does make sense to do that 00:10:26 is...of is not noise. v. useful 00:10:56 Er...it's purely syntactic reversal, isn't it? 00:10:56 yes. v. useful. 00:11:15 I'll willing to be shown cases...:) 00:11:15 :Dan :child :Brennan. :Brennan is :child of :Dan. 00:11:26 er..yes? 00:11:44 actually, that's not as interesting as when it's used in combination with [] 00:11:50 :Brennan isn't a subject isnt it? 00:11:53 is it? 00:11:57 :Class is rdf:type of :MyClass, :AnotherClass, :YetAnotherClass [...] . 00:12:13 I eman, we don't have a new property from is :child of, do we? 00:12:30 no, not a new property. it's just syntactic sugar. but v. handy. 00:12:33 [ daml:inverseOf :child ] . 00:12:53 it's logically consistent with that, sbp, but it doesn't mean that. 00:13:12 yep, I know. That triple isn't in the graph that's read in, it just reverses it 00:13:32 i.e. it's not the case that { :Brennan is :child of :Dan} log:includes { :Dan [ daml:inverseOf :child] :Brennan}. 00:13:36 Right. Well, my resistence is that it screws with read order *and* with number of tokens in a sentences. 00:13:41 * DanC really has to go now. 00:13:58 No biggy for a parser, but for a human reader or someone walking the file with regexps... 00:14:30 walking the file with RegExps: use an N3 parser! 00:14:43 Well, that's the less compelling bit :) 00:14:50 And it's not what *I'd* do :) 00:14:59 but you're special 00:15:11 Exactly, the rexeps aren't for me :) 00:15:43 I'm just not seeing the *very* useful bits. 00:15:44 Ok, 00:15:49 I should do a set of decent RegExps for NTriples, but I'd only end up trying to extend them to URI-views as well... ugh 00:15:59 [is :child of :Dan] is handy. 00:16:14 :Dan :child []. 00:16:18 But *not* handy enough! :) 00:16:33 yes, it's more "handy" than "useful", but that's what N3 is *about* 00:16:50 if you want something that's simple to parse, go use NTriples! 00:17:04 no, I want something easy to author and easy for humans to read. 00:17:27 [ is :child of :Dan ] is pretty easy to read, IMO 00:17:27 I'm not totally against inverses, but I don't like having "symbol" like stuff that don't map to nodes. 00:17:47 yeah... but how else would you do it? 00:18:02 I wouldn't. 00:18:17 >-foo-> and <-foo-< are at least symetricaly. 00:18:25 But horrid. Really bad. Not worth it. 00:18:31 heh: [ dlihc: :Dan ] . 00:18:40 :) 00:18:55 * sbp comes up with some great ideas, sometimes 00:19:05 Alas that's ambiguous. 00:19:10 Or dangerous, at least. 00:19:20 it's pathetic, more than anything :-) 00:19:20 Since you could have a @prefix dlihc: 00:19:40 yeah 00:20:44 'a' at least maps to a uri. 00:20:52 is and of don't. 00:20:55 And they're not operators. 00:20:56 what about "this"? 00:21:01 I don't understand this. 00:21:07 But it looks like it maps to a uri too. 00:21:27 it identifies the root context of the document... rather, the formulae that the document parses to; the semantics as Sandro calls it 00:21:43 Then it's like 'a' 00:22:08 but it doesn't have a URI like "a" and "=" 00:22:16 not <>? 00:22:20 At least in the sense that I see it and can say, "There's some URI it maps to" 00:22:23 It doesn't? 00:22:26 well... CWM used to reserve #_formula 00:22:42 <> means "this document's URI", not "this document's formulae" 00:22:56 you can say it maps to a resource, I guess 00:23:06 which you can't with "is" and "of" 00:23:11 Hmm. I would have though that teh documents uri named the logical formulae... 00:23:25 But a genid approache works for me. 00:23:27 the document's URI names the document! 00:23:56 no, a GenID approach is a terrible idea, because if the document author uses #_formula, it screws it right up 00:24:16 Er.."Magic id" I meant. 00:24:25 cf. http://lists.w3.org/Archives/Public/www-archive/2001Sep/0052 00:24:36 ah... yeah, that's alright 00:24:41 What happens if you concat documents? 00:24:54 forAll and forSome should probably be built in too... 00:25:13 if you concat the *documents*, for "a" and "b" being concated, you get "ab" 00:25:31 viz. [ string:concat ("a" "b"); = "ab" ] . 00:25:48 Er... it's not clear to me if a document establishes a context. 00:25:55 I.e., has an implicit {} around it. 00:26:10 Hmm. No, that's unclear. 00:26:25 That's the concatentation of the documents *text*. 00:26:43 the document's text is synonymous with the document 00:26:53 Er... 00:27:05 No, since distinct documents can have the same text. 00:27:08 Hmm... no, perhaps not 00:27:20 * sbp finds the CWM verbiage for this 00:28:00 try "content" and "semantics" in http://www.w3.org/2000/10/swap/log.n3 00:28:18 if you try to concat two documents, you probably get a category error 00:29:13 ooh: log:conclusion 00:39:05 Well, sean, have you improved my n3parser.pl yet? :) 00:39:49 yes; do you want me to email you the fixes? 00:39:54 Indeed! 00:40:02 just kidding... (as if it wasn't apparent) 00:40:15 oh well, there's always hoping :) 00:40:15 yes, a document has an implicit {} around it. 00:40:26 I might load up SWI-Prolog and test it, though... 00:40:47 Yeek! A lurking lap! 00:41:09 if you do, you need to write a repl. 00:41:44 could you walk me through it? 00:41:47 Or just use read_n3_form(X), statement(C,X []). 00:41:57 At the command line. 00:42:10 That will read and enter one n3 statment. 00:42:29 this doesn't map to a URI. it's magic syntax; it refers to the current scope/document. 00:42:51 To the document's {}, yes? 00:43:28 {this :is :OneThing. { this :is :Another}}. 00:43:41 Er... 00:43:41 Er.. to the enclosing scope. 00:43:43 I.e., a genid? 00:43:54 this is pretty bogus, really. it's only there for the this log:forAll :x construct, which is bogus. 00:44:08 yes, to the enclosing scope, however you implement that. 00:44:17 I'm happy with, this dying ;) 00:44:31 well, you'll have to make up another way of introducing variables. 00:44:58 Yes, I'm trying to figure out variables now. 00:45:10 Since, obviously, the interesting inferences involve them :) 00:48:03 Hmm. /me doesn't understadn teh log:forAll :x construct. 00:48:23 Oh! 00:48:36 this log:forAll :x....got it. 00:49:03 Hmm. 00:49:18 it may be bogus, but it works 00:50:02 Is there any advantage to haveing varialbe declaration be explicit like that? 00:50:05 of course, just as _:x is an exivar, you can have ?x for univars, but then you get the scoping problem 00:50:16 As opposed to ?x style? 00:50:27 yes, for example... 00:50:41 Make it lexically scoped? 00:50:49 actually, Dan did the best example of this. Let me find it... 00:51:02 To the context? 00:52:11 [[[ 00:52:12 consider: (forall (?p) (=> (forall (?x) (likes ?x ?p)) (type ?p GoodGuy)). 00:52:12 i.e. if everybody likes P, then P is a good guy. 00:52:12 that's *very* different from (forall (?p ?x) (=> (likes ?x ?p) (type ?p GoodGuy))) 00:52:12 ]]] 00:53:29 in N3: { this log:forAll :p . { this log:forAll :x . :x :likes :p } log:implies { :p a :GoodGuy } } is different from { this log:forAll :x , :p . { :x :likes :p } log:implies { :p a :GoodGuy } } 00:54:55 Erhm.. how? 00:55:20 so you could replace this forAll :x with ?x, but you'd have to say that ?x is scoped to whatever 00:56:13 how is it different? the former says that if everyone likes :p, :p is a good guy. The latter is if everybody likes everybody, than everybody is a good guy (I think) 00:56:41 Er..Ok, I'm having trouble reading all this, but it seems like the claim is (p)(x)(Lxp -> Gp) is different from (p)((x)(Lxp)->Gp) 00:56:43 er... rather, if anybody_x likes anybody_p, then anybody_p is a good guy 00:57:02 well, I'll clearly have trouble reading your notation too, so that's no help :-) 00:57:17 yeah, that looks about right, though 00:58:33 Hmm. I'm not up on my quantification theorems. 00:58:42 so the scope of the quanitifcation has to be explicit, and more than that, it has to be something that the creator of the rule can specify. If ?x is scoped to the document context at all times, then the former rules becomes impossible 00:58:48 (Dan's former rule) 00:59:14 s/rules/rule/ 00:59:44 er..remember I wanted ?x to be scoped to the immediately enclosing context. 01:01:02 so, what does { ?x :likes ?p } log:implies { ?p a :GoodGuy } represent? how do I represent the other rule? 01:01:06 Well, to something. 01:01:34 you need to be able to specify that something 01:01:53 * bijan 's brain isn't working for this. 01:03:21 Once again: Webize KIF! 01:03:24 :-) 01:04:31 Er...what does kif need to be webized? 01:04:38 What's wrong with it as is? 01:04:39 sbp has quit 01:04:44 sbp has joined #rdfig 01:05:36 it can't be used on the Semantic Web 01:06:10 actually, if you squint right, it can 01:06:16 heh, heh 01:06:28 and if you're willing to write all your URIs out longhand every time. 01:06:41 actually, I think the law of the excluded middle has to go. 01:06:47 well, if I can convert N3 into it... 01:06:59 which I can. see KIFSink.py 01:07:22 but then it's kind of besides the point, which is using KIF as a solid replacement :-) 01:07:49 (I liked the work you did on it to rdf-logic, BTW) 01:08:12 which... the "formal model for scopes" bit? 01:08:30 yeah: http://lists.w3.org/Archives/Public/www-rdf-logic/2001Sep/0004 01:08:38 that goes under the heading of "DanC spends years trying to figure out what's in TimBL's head and relate it to conventional literature" 01:09:07 Poor DanC! 01:09:07 heh! 01:09:12 * DanC_lap wanders off 01:10:25 Hmm. I don't htink I handle uri frags. At all much less correctly. 01:10:29 Cruft cruft cruft. 01:11:15 Hmm. Looks like a tweak to the tokenizer is needed. 01:11:23 what's going wrong, exactly? 01:11:46 Two things. 01:11:56 1) I don't expand <#title> 01:12:07 So my uri validator chokes. 01:12:46 And I need to supply a default URI for the expansion. 01:12:56 use the document URI 01:13:02 What document? 01:13:12 If you're at a REPL, there isn't one :) 01:13:16 either file://, or http://, or wherever you got it from 01:13:27 Didn't get it from no where :) 01:13:29 oh... well, generate a UUID, perhaps? 01:13:33 yep yep. 01:13:37 not a biggy. 01:14:04 hang on... if you're putting in "#title" from a "thing" that has no URI, then that's pretty (to use DanC's word of the day) bogus 01:14:22 what's a relative URI-ref worth when there's nothing relative to it? 01:14:40 well, I wasn't using #title. 01:14:46 I just saw it in Primer. 01:15:17 Actually, i won't have to tweak the tokenizer so long as # only appears inside <> 01:15:20 its ok to use a self-relative URI when you don't know what self is. 01:16:01 well, not in my code, since it'll break the uri check I put in :) 01:16:04 <> a irc:channel. 01:16:21 So, I need to munge something behind the scenes. 01:16:52 Of course, this is, perhaps, not the common case :) 01:17:11 What happens if you parse and output, though? 01:17:42 Me in particular? 01:17:46 then again, you'll be spewing out the same "document"... 01:18:07 nope, taking to myself; just ignore me :-) 01:18:21 oops, prefix checking is...well...not there. 01:19:53 Nor is empty prefixes... dang. 01:20:48 heh! wonderful function names in: http://www.w3.org/2000/10/swap/n-triples2xplanet.pl 01:25:42 GabeW has quit 01:30:55 mnot has quit 01:41:11 timbl has quit 01:41:21 timbl has joined #rdfig 01:41:37 timbl has quit 01:41:44 timbl has joined #rdfig 01:44:24 sbp, are you planning on testing/playing with n3parser.pl now? 01:45:18 no, I'm worrying about hash vs. slash at the moment. I like to sit and worry about things every now and then 01:45:47 Hmm. I don't read slashdot and only occasionally slashdot.... 01:46:11 erm... I meant hash vs. slash in the Semantic Web sense: "to hash, or not to hash" 01:46:27 http://example.org/x vs. http://example.org/#x 01:48:02 * bijan just about to leave, so if you want setup help for tonight... 01:49:38 I'll defer for now :-) 01:49:43 Heh. 01:49:44 Wimp. 01:49:49 heh, heh 01:49:54 well, how long do you have? 01:50:00 A few minuts. 01:50:03 not enough 01:50:07 Er...they were minutes. 01:50:08 Really? 01:50:15 well, you can try. Tell me what to do 01:50:24 You have swi yes? 01:50:31 You have rdf_db, yes? 01:50:32 I've started SWI-Prolog, I have your file on my desktop 01:50:53 rdf_db: dunno, but assume that I have 01:51:07 ['pathto/n3parser.pl']. 01:51:16 (That consults it) 01:51:42 (The other thing you can do is launch XPCE and use PCEmacs. 01:52:23 % n3parser.pl compiled 0.55 sec, 45,812 bytes 01:52:30 You're good. 01:52:32 Type: 01:52:44 * AaronSw keeps thinking .pl -> perl 01:52:56 * sbp too. I use .P 01:52:56 we need namespaced file extensions ;) 01:53:01 heh, heh 01:53:23 read_n3_form(X),rdf_db(Z),statement(Z, X, []). 01:53:24 Ooo 01:53:25 Wait 01:53:32 first, register any namespaces. 01:53:40 or just longer: .prolog, .perl 01:54:07 rdf_register_ns(junk, 'http://foo/#'). 01:54:20 Ok, then the prior line. 01:54:28 Then you can write an N3 sentence, eg., 01:54:33 sbp has quit 01:54:38 sbp has joined #rdfig 01:54:48 junk:ha 'foo'. 01:54:54 (Only single quotes work.) 01:55:17 (Note, up arrow goes through the history, very handy for this sort of thing) 01:55:24 No, 'a' either. 01:55:36 Grpmgprmgrgh, sorry about that... 01:55:46 Action (h for help) ? df_register_ns(junk, 'http://foo/#').Unknown option (h for help) 01:56:00 * bijan 's code is so powerful, it kills sbps! 01:56:08 "a" aborts. 01:56:29 well, at the action prompt. 01:56:45 ERROR: Undefined procedure: df_register_ns/2 01:56:45 Label ERROR not found. 01:56:56 That's becasuse it's rdf_register_ns :) 01:56:57 mnot has joined #rdfig 01:57:03 Not df :) 01:57:04 ah... 01:57:30 ERROR: Undefined procedure: rdf_register_ns/2 01:57:30 Label ERROR not found. 01:57:36 Hmm. 01:57:45 [library(rdf_db)]. 01:58:14 ERROR: source_sink `library(rdf_db)' does not exist 01:58:14 Label ERROR not found. 01:58:24 You need to dl rdf_db. 01:58:32 from? 01:58:33 Here: http://gollem.swi.psy.uva.nl/twiki/pl/bin/view/Code/RdfDB 01:58:46 Sepecifically: http://gollem.swi.psy.uva.nl/twiki/pl/pub/Code/RdfDB/rdf_db.pl 01:58:59 * sbp does so 01:59:42 If you put it in your library folder, [library(rdf_db)]. will work. 01:59:51 Otherwise, ['fullpath']. 01:59:56 it compiled it 02:00:08 Ok, rdf_register_ns/3 02:00:10 er/2 02:01:50 Yes 02:02:14 ?- junk:ha 'foo'. 02:02:14 ERROR: Syntax error: Operator expected 02:02:14 ERROR: junk:ha ERROR: ** here ** 02:02:14 ERROR: ://foo> 'foo' . 02:02:14 Label ERROR not found. 02:02:15 Label ERROR not found. 02:02:16 Label ERROR not found. 02:02:17 Label ERROR not found. 02:02:27 oh, quiet chumpy! 02:02:44 You have to do the read_n3_form(X), rdf_db(C), statement(C, X, []). 02:03:34 hooray! 02:03:41 S: junk:ha 02:03:41 Label S not found. 02:03:41 V: http://foo 02:03:41 O: literal(foo) 02:03:42 Label V not found. 02:03:43 Label O not found. 02:03:47 aaaargh 02:04:05 anyway, it worked :-) 02:05:25 That gives you an n3 line. 02:05:27 Now do, at the ?-, rdf(X,Y,Z). 02:05:27 There you go :) 02:06:15 ooh, cool 02:06:23 X = 'http://foo/#ha' 02:06:23 Y = 'http://foo' 02:06:23 Z = literal("foo") 02:06:35 There you are :) 02:07:09 thank you :-) 02:07:21 well, that gives me something else to hack on when I' 02:07:23 (I have to think about whether the 'literal' is worth keeping from rdf_db since I'm using actuall string objects and can differentiate by tyep. 02:07:36 when I'm bored... an implementation of CWM in Prolog :-) 02:07:42 It does *really* simple (propositional) inference too. 02:08:10 {buncha ground statments} log:implies {buncha ground statments}. 02:08:19 hey, neat 02:08:31 For each of those, it checks to see if hte first batch is in the current scope, if so, it returns the second batch. 02:08:57 The predicate is simple_imply(X,Y). 02:11:27 user5junk:x? 02:11:41 er...some debugging code. 02:11:54 Or, rather, where are you seeing that? 02:11:57 how do I make the inference? I believe I've added the rule 02:12:07 [[[ 02:12:07 | { junk:x junk:y junk:z } log:implies { junk:z junk:y junk:x } . 02:12:07 subject: user5junk:x 02:12:07 S: junk:x 02:12:08 Label S not found. 02:12:09 [...] 02:12:10 ]]] 02:12:16 Yes. 02:12:25 now add the antecedent to the top scope. 02:12:39 (i.e., you won't get nothing if the antecedent isn't available ;)) 02:12:43 then 02:12:53 done... 02:12:56 simple_imply(X,Y). 02:13:06 Should print out the consequent. 02:13:21 heh, great: it did 02:13:30 ?- simple_imply(X,Y). 02:13:30 [http://foo/#z, http://foo/#y, http://foo/#x] 02:13:43 mnot has quit 02:13:45 That's pretty much all there is at the moment :) 02:13:55 well, that's still quite impressive 02:14:18 I need to factor out the lexer and parser into separate moduels for a it all to be "neat". 02:14:22 "01:50:00 A few minuts." you said, quarter of an hour ago :-) 02:14:27 :) 02:14:33 rather, 25 minutes ago... 02:14:47 Oh well. 02:14:55 well, you need it to be better than CWM for it to be really neat 02:15:03 Hmm. No., 02:15:14 clearer than CWM would suffice. 02:15:18 heh, heh! 02:15:20 Faster than CWM would do. 02:15:31 yeah, CWM is not all that fast no complex queries 02:15:36 s/no/on/ 02:15:37 Realize that the REPL coudl be made n3 specific. 02:15:55 Also, I like being able to do direct prolog queries with rdf(...) 02:16:09 Inspecting the 'guts' of an n3 doc seems a bit easier. 02:16:29 timbl has quit 02:16:29 chihchun has quit 02:16:29 bijan has quit 02:16:29 sbp has quit 02:16:29 tav has quit 02:16:29 xena has quit 02:16:29 tav` has quit 02:16:29 Iron_SpermWhale has quit 02:16:29