The Painting Fool - About Me

Hello. I'm the painting fool. I'm a computer program that aspires to be an artist. I've been taught to sketch, draw and paint by my teacher, Dr. Simon Colton, since 2001. I differ from other graphics software by trying to simulate the painting process rather than just the results of the painting process. Painting is a highly cognitive activity which requires skill, appreciation and imagination. Programs such as Photoshop have some skill in being able to rapidly turn a digital photo into an image which looks like it might have been painted in, say, an impressionistic style. But the software is merely a tool to enable humans to be more creative. This is very useful, but Photoshop is not creative, because it is neither appreciative nor imaginative, so it will never be thought of as an artist in its own right. Having said all that, I'm not sure I'm creative myself yet. I've been engaging in a few projects which enable me to express skill, appreciation and imagination, as described below.

Skill

Currently, I mainly work from digital images to produce artworks. My skill lies in being able to look at an image as a collection of paint regions, determine which colours would work for painting the regions, then simulating the usage of all sorts of art materials to produce the picture on a simulated canvas.

During the Workshop on Modular Ontologies at  FOIS 2010, I attended the invited talk by Simon Colton ("Towards Ontology Use, Re-use and Abuse in a Computational Creativity Collective"). He discussed what is probably the most interesting and unique application of AI that I've ever seen: Computational Creativity. He talked about how the turing test was not useful to gauge the effectiveness of an AI to produce creative artifacts. It was the only session I saw in that workshop, but it was one of the most insightful presentations I saw at FOIS.

I had a chance to talk with him later that evening and he pointed me to some slides on Visual Grammars in general and for learning Context Free Design Grammar and the Context Free Art program (which I always wanted to learn but didn't have to time to dig in).

The Painting Fool is a result of the research of him and his team. Some of the work it creates and the parameters they can use to tweak the emotive aspects of the results is incredibly impressive.

GRDDL: A Knowledge Rrepresentation Architectural Style? - part one

Yay. GRDDL is now a W3C Recommendation!

I'm very proud to be a part of that and there is alot about this particular architectural style that I have always wanted to write about. I recently came upon the opportunity to consider one particular facet.

This is why it seems the same as with GRDDL. There are tranformations you can make, but they are not entailments in a logic, they just go from one graph to a different graph.

Yes, that is one part of the larger framework that is well considered. GRDDL does not rely on logical entaiment for its normative definition. It is defined operationally, but can also be described via declarative (formal) semantics. It defines a mapping (not a function in the true sense - the specification clearly identifies ambiguity at the level of the infoset) from an XML representation of an "information resource" to a typed RDF representation of the same "information resource". The output is required to have a well-defined mapping of its own into the RDF abstract syntax.

The less formal definition uses a dialect of Notation 3 that is a bit more expressive than Datalog Logic Programming (it uses function symbols - builtins - in some of the clauses ). The proof at the bottom of that page justifies the assertion that http://www.w3.org/2001/sw/grddl-wg/td/titleauthor.html has a GRDDL result which is composed entirely of the following RDF statement:

<http://musicbrainz.org/mm-2.1/album/6b050dcf-7ab1-456d-9e1b-c3c41c18eed2> is named "Are You Experienced?" .

Frankly, I would have gone with "Bold as Love", myself =)

Once you have a (mostly) well-defined function for rendering RDF from information resources, you enable the deployment of useful ( and re-usable ) interpretations for intelligent agents (more on these later). For example, the test suite, is a large semantic web of XML documents that GRDDL-aware agents can traverse, performing Quality Assurance tests (using EARL) of their conformance to the operational semantics of GRDDL.

However, it was very important to leave entailment out of the equation until it serves a justifiable purpose. For example, a well-designed RDF querying language does not require logical entailment (RDF, RDFS, OWL, or otherwise) for it to be useful in the general case. You can calculate a closure (or Herbrand base) and then dispatch structural graph queries. This was always true with Versa. You can glean (pun intended) quite a bit from only the structural nature of a Graph. A whole generation of graph theoretical literature demonstrates this.

In addition, once you have a well-defined set of semantics for identifying information resources with RDF assertions that are (logically) relevant to the closure, you have a clear seperation between manipulation of surface syntax and full-blown logical reasoning systems.

It should be considered a semantic web architectural style (if you will) to constrain the use of entailment to only where it has some demonstrated value to the problem space. Where it makes sense to use entailment, however, you will find the representations are well-engineered for the task.

Chimezie Ogbuji

via Copia

Prooving a URI is a Document (Formally)

I'm about ready to check-in some updates for proof generation in FuXi, and was having some fun with httpRange-14 I wanted to share. Some time ago, I wrote Emeka, a phenny-based IRC bot to experiment with how intelligent agents might best be used over IRC. I've recently been playing around with using REGEX patterns to match and invoke "Semantic Web" services (none seem "generally" useful, so far). But, I recently added a "hook" to generate proofs of RDF assertions from rulesets and ontologies. Below is how Emeka & FuXi deduce that http://dbpedia.org/data/DBpedia is a "document" as defined by the FOAF ontology:

<chimezie> .varBind proofSrc list('http://dbpedia.org/data/DBpedia','http://xmlns.com/foaf/spec/')
<chimezie> Emeka, proove "@prefix foaf: <http://xmlns.com/foaf/0.1/>.  <http://en.wikipedia.org/wiki/DBpedia> a foaf:Document ."  from $proofSrc
<Emeka>  Calculated closure in 514.962911606 milli seconds
<Emeka>  reached Q.E.D. See: http://rafb.net/p/5PvAhz98.html

The first line uses a Versa expression to build a variable bound to a list of resources, which can be referenced within Versa expressions (these are the ontologies that should be used with "graph links"). The second line uses a combination of Notation 3 (to express the "goal") and Versa (for the list of information resources). I'm still considering whether having an IRC bot automatically paste content to a "paste bin" is rude.

The contents of the pastebin comprises a (somewhat) human-readable (linear) proof.

Human-readable proof:
Building nodeset around antecedent goal (justified by a direct assertion): foaf:page(dbpedia:DBpedia wiki:DBpedia)
Building inference step for Proof step for foaf:topic(wiki:DBpedia dbpedia:DBpedia)
Inferred from RETE node via foaf:topic(?X ?sgflxxAo413) :- foaf:page(?sgflxxAo413 ?X)
Bindings: {?X: rdflib.URIRef('http://en.wikipedia.org/wiki/DBpedia'), ?sgflxxAo413: rdflib.URIRef('http://dbpedia.org/resource/DBpedia')}
Building nodeset around antecedent goal: foaf:topic(wiki:DBpedia dbpedia:DBpedia)
Building inference step for Proof step for foaf:Document(wiki:DBpedia)
Inferred from RETE node via foaf:Document(?X) :- foaf:topic(?X ?sgflxxAo1286)
Bindings: {?X: rdflib.URIRef('http://en.wikipedia.org/wiki/DBpedia'), ?sgflxxAo1286: rdflib.URIRef('http://dbpedia.org/resource/DBpedia')}
Building proof around goal: foaf:Document(wiki:DBpedia)
## Generated by Emeka at 2007-09-12T20:52:53.728814 in 6.64019584656 milli seconds

Note, the proof doesn't rely on what the transport protocol says about http://en.wikipedia.org/wiki/DBpedia

I had some problems getting the Standford Inference Web proof browser to render FuXi's RDF/XML serialization, I wonder if it is because of my assumption (in the serialization) that the 'entailment' semantics of a RETE network is equivalent to Generalized Modus Ponens. Anyway, the diagram FuXi generates is below:

Proof about DBPedia

This was generated via:

Fuxi --output=dot --ns=wiki=http://en.wikipedia.org/wiki/  --ns=dbpedia=http://dbpedia.org/resource/  --ns=foaf=http://xmlns.com/foaf/0.1/ --proove="@prefix foaf: <http://xmlns.com/foaf/0.1/>.  <http://en.wikipedia.org/wiki/DBpedia> a foaf:Document ." --dlp http://xmlns.com/foaf/spec/ http://dbpedia.org/data/DBpedia

I just need to wrap up the doctests and unit tests and push the latest build to cheeseshop and google code

Chimezie Ogbuji

via Copia

FuXi: Becoming a Full-fledged Logical Reasoning System

[by Chimezie Ogbuji]

I've been doing alot of "Google reading" lately

Completing Logical Reasoning System Capabilities

With the completion (or near completion) of PML-generating capabilities for FuXi, it is becoming a fully functional logical reasoning system. In "(Artificial Intelligence: A Modern Approach)" Stuart Russel and Peter Norvig identify the following main categories for automated reasoning systems:

  1. Theorem provers
  2. Production systems
  3. Frame systems and semantic networks
  4. Description Logic systems

OWL and RDF are coverage for 3 and 4. The second category is functionally covered by the RETE-UL algorithm FuXi employs (a highly efficient modification of the original RETE algorithm). The currently developing RIF Basic Logic Dialect covers 2 - 4. Proof Markup Language covers 1. Now, FuXi can generate (and export visualization diagrams) Proof Markup Language (PML) structures. I still need to do more testing, and hope to be able to generate proofs for each of the OWL tests. Until then, below is a diagram of the proof tree generated from the "She's a Witch and I have Proof" test case:

# http://clarkparsia.com/weblog/2007/01/02/burn-the-witch/
# http://www.netfunny.com/rhf/jokes/90q4/burnher.html
@prefix : <http://www.w3.org/2000/10/swap/test/reason/witch#>.
@keywords is, of, a.
#[1]    BURNS(x) /\ WOMAN(x)            =>      WITCH(x)
{ ?x a BURNS. ?x a WOMAN } => { ?x a WITCH }.
#[2]    WOMAN(GIRL)
GIRL a WOMAN.
#[3]    \forall x, ISMADEOFWOOD(x)      =>      BURNS(x)
{ ?x a ISMADEOFWOOD. } => { ?x a BURNS. }.
#[4]    \forall x, FLOATS(x)            =>      ISMADEOFWOOD(x)
{ ?x a FLOATS } => { ?x a ISMADEOFWOOD }.
#[5]    FLOATS(DUCK)
DUCK a FLOATS.
#[6]    \forall x,y FLOATS(x) /\ SAMEWEIGHT(x,y) =>     FLOATS(y)
{ ?x a FLOATS. ?x SAMEWEIGHT ?y } => { ?y a FLOATS }.
# and, by experiment
# [7]   SAMEWEIGHT(DUCK,GIRL)

Shes a witch and I have proof trace

There is another test case of the "Dan's home region is Texas" test case on a python-dlp Wiki: DanHomeProof:

@prefix : <gmpbnode#>.
@keywords is, of, a.
dan home [ in Texas ].
{ ?WHO home ?WHERE.
  ?WHERE in ?REGION } => { ?WHO homeRegion ?REGION }.

Dans home region is Texas proof

I decided to use PML structures since there are a slew of Stanford tools which understand / visualize it and I can generate other formats from this common structure (including the CWM reason.n3 vocabulary). Personally, I prefer the proof visualization to the typically verbose step-based Q.E.D. proof.

Update: I found nice write-up on the CWM-based reason ontology and translations to PML

So, how does a forward-chaining production rule system generate proofs that are really meant for backward chaining algorithms? When the FuXi network is fed initial assertions, it is told what the 'goal' is. The goal is a single RDF statement which is being prooved. When the forward-chaining results in a inferred triple which matches the goal, it terminates the RETE algorithm. So, depending on the order of the rules and the order that the initial facts are fed it will be (for the general cases) less efficient than a backward chaining algorithm. However, I'm hoping the blinding speed of the fully hashed RETE-UL algorithm makes up the difference.

I've been spending quite a bit of time on FuXi mainly because I am interested in empirical evidence which supports a school of thought which claims that Description Logic based inference (Tableaux-based inference) will never scale as well the Logic Programming equivalent - at least for certain expressive fragments of Description Logic (I say expressive because even given the things you cannot express in this subset of OWL-DL there is much more in Horn Normal Form (and Datalog) that you cannot express even in the underlying DL for OWL 1.1). The genesis of this is a paper I read, which lays out the theory, but there was no practice to support the claims at the time (at least that I knew of). If you are interested in the details, the paper is "Description Logic Programs: Combining Logic Programs with Description Logic" and written by many people who are working in the Rule Interchange Format Working Group.

It is not light reading, but is complementary to some of Bijan's recent posts about DL-safe rules and SWRL.

A follow-up is a paper called "A Realistic Architecture for the Semantic Web" which builds on the DLP paper and makes claims that the current OWL (Description Logic-based) Semantic Web inference stack is problematic and should instead be stacked ontop of Logic Programming since Logic Programming algorithm has a much richer and pervasively deployed history (all modern relational databases, prolog, etc..)

The arguments seem sound to me, so I've essentially been building up FuXi to implement that vision (especially since it employes - arguably - the most efficient Production Rule inference algorithm). The final piece was a fully-functional implementation of the Description Horn Logic algorithm. Why is this important? The short of it is that the DLP paper outlines an algorithm which takes a (constrained) set of Description Logic expressions and converts them to 'pure' rules.

Normally, Logic Programming N3 implementations pass the OWL tests by using a generic ruleset which captures a subset of the OWL DL semantics. The most common one is owl-rules.n3. DLP flips the script by generating a rule-set specifically for the original DL, instead of feeding OWL expressions into the same network. This allows the RETE-UL algorithm to create an even more efficient network since it will be tailored to the specific bits of OWL.

For instance, where I used to run through the OWL tests in about 4 seconds, I can now pass them in about 1 secound using. Before I would setup a RETE network which consisted of the generic ruleset once and run the tests through it (resetting it each time). Now, for each test, I create a custom network, evaluate the OWL test case against it. Even with this extra overhead, it is still 4 times faster! The custom network is trivial in most cases.

Ultimately I would like to be able to use FuXi for generic "Semantic Web" agent machinery and perhaps even to try some that programming-by-proof thing that Dijkstra was talking about.

Chimezie Ogbuji

via Copia

Where does the Semantic Web converge with the Computerized Patient Record?

I've been thinking alot about the "Computer-based Patient Record: CPR", an acronym as unlikely as GRDDL but once again, a methodology expressed as an engineering specification. In both cases, the methodology is a mouthful, but a coherent architectural "style" and requires a mouthful of words to describe. Other examples of this:

  • Representation State Transfer
  • Rich Web Application Backplane
  • Problem-oriented Medical Record
  • Gleaning Resource Descriptions from Dialects of Languages

The term itself was coined (I think) by the Institute of Medicine [1]. If you are in healthcare and are motivated by the notion of using technology to make healthcare effective and inexpensive as possible, you should do the Institute a favor and buy the book:

National Institutute of Medicine, The Computer-Based Patient Record: An Essential Technology for Health Care - Revised Edition., 1998, ISBN: 0309055326.

I've written some recent slides that are on the W3C ESW 'wiki' which all have something to do with the idea in one way or another:

The nice thing about working in a W3C Interest Group is that the work you do is for the general publics benefit, so it is a manefestation of the W3C notion of the Semantic Web, which primarily involves a human social process.

Sorta like a technological manefestation of our natural darwinian instinct.

That's how I think of the Semantic Web, anyways: as a very old, living thread of advancements in Knowledge Representation which intersected with an anthropological assesment of some recent web architecture engineering standards.

Technology is our greatest contribution and so it sohould only make sense that wherer we use it to better our health it should not come as a cost to us. The slides reference and include a suggested OWL-sanctioned vocabulary for basically implementing the Problem-oriented Medical Record (a clinical methodology for problem solving).

I think the idea of a free (as in beer) vocabulary for people who need healthcare has an interesting intersection with the pragmatic parts of the Semantic Web (avoiding the double quotes) vision. I have exercised-induced asthma (or was "diagnosed" as such when I was younger). I still ran Track-and-Field in Highschool and was okay after an initial period where my lungs had to work overtime. I wouldn't mind hosting RDF content about such a "finding" if it was for my person benefit that a piece of software could do something useful for me in an automated, deterministic way.

"HL7 CDA" seems to be a freely avaiable, well-organized vocabulary for describing messages dispatched between hospital systems. And I recently wrote a set of XSLT templates which extract predicate logic statemnts about a CDA document using the POMR ontology and the other freely available "foundational ontologies" it coordinates. The CDA document on xml.coverpages.org has a nice concise description of the technological merits of HL7 CDA:

The HL7 Clinical Document Architecture is an XML-based document markup standard that specifies the structure and semantics of clinical documents for the purpose of exchange. Known earlier as the Patient Record Architecture (PRA), CDA "provides an exchange model for clinical documents such as discharge summaries and progress notes, and brings the healthcare industry closer to the realization of an electronic medical record. By leveraging the use of XML, the HL7 Reference Information Model (RIM) and coded vocabularies, the CDA makes documents both machine-readable (so they are easily parsed and processed electronically) and human-readable so they can be easily retrieved and used by the people who need them. CDA documents can be displayed using XML-aware Web browsers or wireless applications such as cell phones..."

The HL7 CDA was designed to "give priority to delivery of patient care. It provides cost effective implementation across as wide a spectrum of systems as possible. It supports exchange of human-readable documents between users, including those with different levels of technical sophistication, and promotes longevity of all information encoded according to this architecture. CDA enables a wide range of post-exchange processing applications and is compatible with a wide range of document creation applications."

A CDA document is a defined and complete information object that can exist outside of a messaging context and/or can be a MIME-encoded payload within an HL7 message; thus, the CDA complements HL7 messaging specifications.

If I could put up a CDA document describing the aspects of my medical history that were in my benefit to be freely available (at my discretion), I would do so in the event some piece of software could do some automated things for my benefit. Leveraging a vocabulary which essentially grounds an expressive variant of predicate logic in a transport protocol makes the chances that this happens, very likely. The effect is as multiplicative as the human population.

The CPR specification is also very well engineered and much ahead of its time (it was written about 15 years ago). The only technological checkmark left is a uniform vocabulary. Consensus stands in the way of uniformity, so some group of people need to be thinking about how the "pragmatic" and anthropological notions of the Semantic Web can be realized with a vocabulary about our personally controlled, public clinical content. Don't you think?

I was able to register the /cpr top level PURL domain and the URL http://purl.org/cpr/1.0/problem-oriented-medical-record.owl# resolves to the OWL ontology with commented imports to other very relevant OWL ontologies. Once I see a pragmatic demonstration of leaving owl:imports in a 'live' URL, I'll remove them. It would be a shame if any Semantic Web vocabulary terms came in conflict with a legal mandate which controlled the use of a vocabulary.

Chimezie Ogbuji

via Copia

Progress on two Reference Implementations for RETE and GRDDL

Whew! During the moments I was able to spare while at ISWC (I was only there on monday and tuesday, unfortunately), I finished up two 'reference' implementations I've been enthusiastically hacking on quite recently. By reference implementation, I mean an implementation that attempts to follow a specification verbatim as an exercise to get a better understanding of it.

I still need a write up on the engaging conversations I had at ISWC (it was really an impressive conference, even from just the 2 days worth I got to see) as well as the presentation I gave on using GRDDL and Web architecture to meet the requirements of Computer-based Patient Records.

FuXi and DLP

The first milestone was with FuXi, which I ended up rewriting completely based on some exchanges with Peter Lin and Charles Young.

This has probably been the most challenging piece of software I've ever written and I was quite niave in the beginning about my level of understanding of the nuances of RETE. Anyone interested in the formal intersection of Notation 3 / RDF syntax and the RETE algorithm(s) will find the exchanges in the comments of the above post very instructive - and Peter Lin's blog in general. Though he and I have our differences in the value of mapping RETE to RDF/N3 his insights into my efforts have been very helpful.

In the process, I discovered Robert Doorenbos PhD thesis "Production Matching for Large Learning Systems" which was incredibly valuable in giving me a comprehensive picture of how RETE (and RETE/UL) could be 'ported' to accomodate Notation 3 reasoning.

The primary motivation (which has led to many sleepless nights and late night hackery) is what I see as a lack of understanding within the community of semantic web developers of the limitations of Tableux-based reasoning and the common misconception that the major Tableaux-based reasoners (Fact++, Racer, Pellet, etc..) represent the ceiling of DL reasoning capability.

The reality is that logic programming has been around the block much longer than DL and has much more mature algorithms available (the primary one being RETE). I've invested quite a bit of effort in what I believe will (eventually) demonstrate very large-scale DL reasoning performance that will put Tableaux-based reasoning to shame - if only to make it clear that more investigation into the intersection of LP and DL is crucial for making the dream of the SW a verbatim reality.

Ofcouse, there is a limit to what aspects of DL can be expressed as LP rules (this subset is called Description Logic Programming). The 'original' DLP paper does well to explain this limitation, but I believe this subset represents the more commonly used portions of OWL and the portions of OWL 1.0 (and OWL 1.1 for that matter) left out by such an intersection will not be missed.
Ivan Herman pointed me to a paper by Horst, Herman which is quite comprehensive in outlining how this explicit intersection can be expressed axiomatically and the computational consequences of such an axiomatization. Ivan used this a guide for his RDFSClosure module.

Not enough (IMHO) has been done to explore this intersection because people are comfy with the confines of non-LP algorithms. The trail (currently somewhat cold) left by the Mindswap work on Pychinko needs to be picked up, followed and improved.

So, I rolled up my sleeves, dug deep and did my best to familiarize myself with the nuances of production system optimization. Most of the hard work has already been done, thanks to Robert Doorenbos subsetting (and extension) of the original Charles Forgy algorithm. FuXi, gets through a large majority the OWL tests using a ruleset that closely implements what Horst lays out in his paper and does so with impressive times - even with more optimizations pending.

The most recent changes include a command-line interface for launching it:

chimezie@Zion:~/devel/Fuxi$ python Fuxi.py --out=n3 --ruleFacts
--ns=owl=http://www.w3.org/2002/07/owl#
--ns=test=http://metacognition.info/FuXi/DL-SHIOF-test.n3#
--rules=test/DL-SHIOF-test.n3
Time to build production rule (RDFLib): 0.0172629356384 seconds
Time to calculate closure on working memory: 224.906921387 m seconds

@prefix owl: <http://www.w3.org/2002/07/owl#>.
@prefix rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#>.
@prefix test: <http://metacognition.info/FuXi/DL-SHIOF-test.n3#>.

 test:Animal test:in _:XRfZIlKy56,
        _:XRfZIlKy57.

 test:Cosi_fan_tutte a test:DaPonteOperaOfMozart;
    test:in _:XRfZIlKy47,
        _:XRfZIlKy48,
        [].

 test:Don_Giovanni a test:DaPonteOperaOfMozart;
    test:in _:XRfZIlKy47,
        _:XRfZIlKy48.

 .... snip ...

FuXi still doesn't support 'built-ins' (or custom comparisons), but luckily the thesis includes a section on how to implement non equality testing of rule constraints that should be (relatively) easy to add. The theseis also includes a section on how negated conditions can be implemented (which is probably the most glaring axiom missing from DLP). Finally, Robert's paper includes a clever mechanism for hashing the Alpha network that has yet to be implemented (and is simple enough to implement) that should contribute significant performance gains.

There are other pleasant surprises with the current codebase. The rule compiler can be used to identify inefficencies in rule patterns, the command-line program can be used to serialize the closure delta (i.e., only the triples inferred from the ruleset and facts), and (my favorite) a Notation 3 ruleset can be exported as a graphviz diagram in order to visualize the rule network. Having 'browsed' various rule-sets in this way, I must say it helps in understanding the nuances of optimization when you can see the discrimination network that the triples are propagated through.

I don't have a permanent home for FuXi yet but have applied for a sourceforge project (especially since SF now supports SVN, apparently). So, until then, FuXi can be downloaded from:

GRDDL Client for 4Suite and RDFLib

During the same period, I've also been working on a 'reference' implementation of GRDDL (per the recently released Working Draft) for 4Suite and RDFLib. It's a bit ironic in doing so, since the 4Suite repository framework has essentially been using GRDDL-like Content Management Systems mechanisms since its inception (sometime in 2001).
However, I thought doing so would be the perfect oppurtunity to:

  • Demonstrate how 4Suite can be used with RDFLib (as a prep to the pending deprecation of 4Suite RDF for RDFLib)
  • Build a framework to compose illustrative test cases for the Working Group (of which I'm a member)
  • As a way to completely familiarize myself with the various GRDDL mechanisms

I posted this to the Working Group mailing list and plan to continue working on it. In particular, the nice thing about the convergence of these two projects of mine is that I've been able to think a bit about how both GRDDL and Fuxi could be used to implement efficient, opportunistic programs that extract RDF via GRDDL and explicit links (rdf:seeAlso, owl:imports, rdfs:isDefinedBy) and perform incremental web closures by adding the triples discovered in this way one at a time to a live RETE network.

The RETE algorithm is tailored specifically to behave as a black box to changes to a production system and so crawling the web, extracting RDF (a triple at a time) and reasoning over it in this way (the ultimate semantic web scenario) becomes a very real scenario with sub-second response times to boot. At the very least, reasoning should cease to be much of a bottleneck compared to the actual dereferencing and parsing of RDF from distributed locations. Very exciting prospects. Watch this space..

Chimezie Ogbuji

via Copia

Rete-inspired N3 Rule Network Finished

See: previous

I called it quits (for now) in trying to retrofit the RETE-based algorithm to handle N3 functions & filters. I've checked in a rewritten Rete module with tests for SHIOF Description Logic axioms.

Tracking dependencies between filter / function variables, became hairy.

An SVG diagram of these rules compiled into a RETE-like network was generated using an included Boost Graph Library utility function.

It's unit test output:

testRules (__main__.TestEvaluateNetwork) ... Time to build production rule (RDFLib): 0.0118989944458 seconds
Time to calculate closure on working memory: 0.478726148605 seconds
ok
----------------------------------------------------------------------
Ran 1 test in 0.751s
OK

I also integrated Python iterator algebra implementation of a relational hash join (for the Beta Nodes).

It workes with RDFLib, so I'd eventually like to integrate the ability to dispatch SPARQL queries over an N3 Closure Graph - generated from the network. The speed in which it is able to render rules graphs rendered the option of serializing a compiled network into persistence a non-issue.

From the README.txt:

Fu Xi (pronounced foo-see) is a forward-chaining inferencing expert system for RDFLib Notation 3 graphs. It is named after the first mythical soveriegn of ancient china who supposedly, 'looked upward and contemplated the images in the heavens, and looked downward and contemplated the occurrences on earth.'.

Originally, it was an idea to express the formalisms of the Yi Jing / I Ching in Description & First Order Logic.

It relies on Charles Forgy's Rete algorithm for the many pattern/many object match problem - which builds a triple pattern reasoning network. It can be used as a reasoner with capabilities for certain expressive Description Logics (via OWL/RDFS axioms in N3) or a general N3 production / expert system. It uses Python hash / set / list idioms to maximize matching efficiency of the network. In particular, it uses an Iterator algebra implementation for the join activation mechanism

An example of its use:

from FuXi.Rete.Util import generateTokenSet
        from FuXi.Rete import *
        from rdflib.Graph import Graph,ReadOnlyGraphAggregate
        from rdflib import plugin
        from FuXi.Rete.RuleStore import N3RuleStore
        store = plugin.get('IOMemory',Store)()
        store.open('')
        ruleGraph = Graph(N3RuleStore())
        ruleGraph.parse(open('some-rule-file.n3'),format='n3')             
        factGraph = Graph(store)
        factGraph.parse(open('fact-file'),format='..')
        closureDeltaGraph = Graph(store)            
        network = ReteNetwork(ruleStore,
                              initialWorkingMemory=generateTokenSet(factGraph),
                              inferredTarget = closureDeltaGraph)            
        for s,p,o in network.closureGraph(factGraph):
            .. do something with triple ..

To check out from cvs:

cvs -d:pserver:anonymous@cvs.4suite.org:/var/local/cvsroot login
cvs -d:pserver:anonymous@cvs.4suite.org:/var/local/cvsroot get Fuxi

Chimezie Ogbuji

via Copia

What Do Closed Systems Have to Gain From SW Technologies?

Aaron Swartz asked that I elaborate on a topic that is dear to me and I didn't think a blog comment would do it justice, so here we are :)

The question is what do single-purpose (closed) databases have to gain from SW technologies. I think the most important misconception to clear up first is the idea that XML and Semantic Web technologies are mutually exclusive.
They are most certainly not.

It's not that I think Aaron shares this misconception, but I think that the main reason why the alternative approach to applying SW technologies that he suggests isn't very well spoken for is that quite a few on the opposing sides of the issue assume that XML (and it's whole strata of protocols and standards) and RDF/OWL (the traditionally celebrated components of SW) are mutually exclusive. There are other misconceptions that hamper this discussion, such as the assumption that the SW is an all or nothing proposition, but that is a whole other thread :)

As we evolve towards a civilization where the value in information and it's synthesis is of increasing importance, 'traditional' data mining, expressiveness of representation, and portability become more important for most databases (single-purpose or not).

These are areas that these technologies are meant to address, explicitly because “standard database” software / technologies are simply not well suited for these specific requirements. Not all databases are alike and so it follows that not all databases will have these requirements: consider databases where the primary purpose is the management of financial transactions.

Money is money, arithmetic is arithmetic, and the domain of money exchange and management for the most part is static and traditional / standard database technologies will suffice. Sure, it may be useful to be able to export a bank statement in a portable (perhaps XML-based) format, but inevitably the value in using SW-related technologies is very minimal.

Ofcourse, you could argue that online banking systems have a lot to gain from these technologies, but the example was of pure transactional management, the portal that manages the social aspects of money management is a layer on top.

However, where there is a need to leverage:

  • More expressive mechanisms for data collection (think XForms)
  • (Somewhat) unambiguous interpretation of content (think FOL and DL)
  • Expressive data mining (think RDF querying languages)
  • Portable message / document formats (think XML)
  • Data manipulation (think XSLT)
  • Consistent addressing of distributed resources (think URLs)
  • General automation of data management (think Document Definitions and GRDDL)

These technologies will have an impact on how things are done. It's worth noting that these needs aren't restricted to distributed databases (which is the other assumption about the Semantic Web - that it only applies within the context of the 'Web'). Consider the Wiki example and the advantages that Semantic Wikis have over them:

  • Much Improved possibility of data mining from more formal representation of content
  • 'Out-of-the-box' interoperability with tools that speak in SW dialects
  • Possibility of certain amount of automation from the capabilities that interpretation bring

It's also worth noting that recently the Semantic Wiki project introduced mechanisms for using other vocabularies for 'marking-up' content (FOAF being the primary vocabulary highlighted).

It's dually important in that 1) it demonstrates the value in incorporating well-established vocabularies with relative ease and 2) the policed way in which these additional vocabularies can be used demonstrate precisely the middle ground between a very liberal, open world assumption, approach to distributed data in the SW and controlled, closed, (single-purpose) systems approach.

Such constraints can allow for some level of uniformity that can have very important consequences in very different areas: XML as a messaging interlingua and extraction of RDF.

Consider the value in developing a closed vocabulary with it's semantics spelled out very unambiguously in RDF/RDFS/OWL and a uniform XML representation of it's instances with an accompanying XSLT transform (something the AtomOWL project is attempting to achieve).

What do you gain? For one thing, XForms-based data entry for the uniform XML instances and a direct, (relatively) unambiguous mapping to a more formal representation model – each of which have their own very long list of advantages they bring by themselves much less in tandem!

Stand-alone databases (where their needs intersect with the value in SW technologies) stand to gain: Portable, declarative data entry mechanisms, interoperability, much improved capabilities for interpretation and synthesis of existing information, increased automation of data management (by closing the system certain operations become much more predictable), and the additional possibilities for alternative reasoning heuristics that take advantage of closed world assumptions.

Chimezie Ogbuji

via Copia

Mapping Rete algorithm to FOL and then to RDF/N3

Well, I was hoping to hold off on the ongoing work I've doing with FuXi, until I could get a decent test suite working, but I've been engaged in several threads (older) that left wanting to elaborate a bit.

There is already a well established precedent with Python/N3/RDF reasoners (Euler, CWM, and Pychinko). FuXi used to rely on Pychinko, but I decided to write a Rete implementation for N3/RDF from scratch - trying to leverage the host language idioms (hashing, mappings, containers, etc..) as much as possible for areas where it could make a difference in rule evaluation and compilation.

What I have so far is more Rete-based than a pure Rete implementation, but the difference comes mostly from the impedance between the representation components in the original algorithm (which are very influenced by FOL and Knowledge Representation in general) and those in the semantic web technology stack.

Often with RDF/OWL, there is more talk than neccessary, so I'll get right to the meat of the semantic mapping I've been using. This assumes some familiarity with the original algorithm. Some references:

Tokens

The working memory of the network is fed by an N3 graph. Tokens represent the propagation of RDF triples (no variables or formulae identifiers) from the source graph through the rule network. These can represent token addition (where the triples are added to the graph - in which case the live network can be associated with a live RDF graph) and token removals (where triples are removed from the source graph). When tokens pass an alpha nodes intra-element test (see below) the tokens are passed on with a subtitution / mapping of variables in the pattern to the corresponding terms in the triples. This variable substitution is used to check for consistent variable bindings across beta nodes.

ObjectType Nodes and Working Memory

ObjectType nodes can be considered equivalent to the test for concept subsumption (in Description Logics). and therefore equivalent to the alpha node RDF pattern:

?member rdf:type ?klass.

'Classic' Alpha node patterns (the one below is taken directly from the original paper) map to multiple RDF-triple alpha node patterns:

(Expression ^Op X ^Arg2 Y)

Would be equivalent to the following triples patterns (the multiplicative factor is that RDF assertions are limited to binary predicates):

  • ?Obj rdf:type Expression
  • ?Obj Op ?X
  • ?Obj Arg2 ?Y

Alpha Nodes

Alpha nodes correspond to patterns in rules, which can be

  1. Triple patterns in N3 rules
  2. N-array functions.

Alpha node intra-element tests have a 'default' mechanism for matching triple patterns or they exhibit the behavior associated with a registered set of N-array functions - the core set coincide with those used by CWM/Euler/Pychinko (and often called N3 built-ins). Fuxi will support an extension mechanism for registering additional N-array N3 functions by name which associate them with a Python function that implements the constraint in a similar fashion to SPARQL extension functions. N-aray functions are automatically propagated through the network so they can participate in Beta Node activation (inter-element testing in Rete parlance) with regular triple patterns, using the bindings to determine the arguments for the functions.

The default mechanism is for equality of non-variable terms (URIs).

Beta Nodes

Beta nodes are pretty much verbatim Rete, in that they check for consistent variable substitution between their left and right memory.This can be considered similar to the unification routine common to both forward and backward chaining algorithms which make use of the Generalized Modus Ponens rule. The difference is that the sentences aren't being make to look the same but the existing variable substitutions are checked for consistency. Perhaps there is some merit in this similarity that would make using a Rete network to faciliate backward-chaining and proof generation an interesting possiblity, but that has yet to be seen.

Terminal Nodes

These correspond with the end of the LHS of a N3 rule, and is associated with the RHS and when 'activated' they 'fire' the rules, apply the propaged variable substitution, and add the newly inferred triples to the network and to the source graph.

Testing and Visualizing RDF/N3 Rete Networks

I've been able to adequately test the compilation process (the first of two parts in the original algorithm), using a visual aid. I've been developing a library for generating Boost Graph Library DiGraphs from RDFLib graphs - called Kaleidos. The value being in generating GraphViz diagrams, as well as access to a whole slew of graph heuristics / algorithms that could be infinitely useful for RDF graph analysis and N3 rule network analysis:

  • Breadth First Search
  • Depth First Search
  • Uniform Cost Search
  • Dijkstra's Shortest Paths
  • Bellman-Ford Shortest Paths
  • Johnson's All-Pairs Shortest Paths
  • Kruskal's Minimum Spanning Tree
  • Prim's Minimum Spanning Tree
  • Connected Components
  • Strongly Connected Components
  • Dynamic Connected Components (using Disjoint Sets)
  • Topological Sort
  • Transpose
  • Reverse Cuthill Mckee Ordering
  • Smallest Last Vertex Ordering
  • Sequential Vertex Coloring

Using Kaleidos, I'm able to generate visual diagrams of Rete networks compiled from RDF/OWL/N3 rule sets.

However, the heavy cost with using BGL is the compilation process of BGL and BGL python, which is involved if doing so from source.

Chimezie Ogbuji

via Copia

Practical Temporal Reasoning with Notation 3

Recently, Dan Brickley expressed an interest in the extent to which Bioinformatic research efforts are leveraging RDF for temporal reasoning (and patient healthcare record integration - in general). The thread on the value of modeling temporal relations explicitly versus relying on them being built into core RDF semantics left me feeling like a concrete example was in order.

We have a large (3500+ assertions) OWL Full ontology describing all the data we collect about Cardiothoracic procedures (the primary purpose of our database as currently constituted – in a relational model). There are several high-level classes we use to model concepts that, though core to our model, can be thought of as general enough for a common upper ontology for patient data.

One of the classes is ptrec:TemporalData (from here on out, I'll be using the ptrec prefix to describe vocabulary terms in our ontology) which is the ancestor of all classes that are expressed on an axis of time. We achieve a level of precision in modeling data on a temporal axis that enhances the kind of statistical analysis we perform on a daily basis.

In particular we use three variables:

  • ptrec:startDT (xs:dateTime)
  • ptrec:stopDT (xs:dateTime)
  • ptrec:instantDT (xs:dateTime)

The first two are used to describe an explicit (and 'proper') interval for an event in a patient record. This is often the case where the event in question only had a date associated with it. The latter variable is used when the event is instantaneous and the associated date / time is known.

The biggest challenge isn't simply the importance of time in asking questions of our data but of temporal factors that are keyed off specific, moving points of reference. For example, consider a case study on the effects of administering a medication within X days of specific procedure. The qualifying procedure is key to the observations we wish to make and behaves as a temporal anchor. Another case study interested in the effects of administering the same medication but with respect to a different procedure should be expected to rely on the same temporal logic – but keyed off a different point in time. However, by being explicit about how we place temporal data on a time axis (as instants or intervals) we can outline a logic for general temporal reasoning that can be used by either case study.

Linking into an OWL time ontology we can setup some simple Notation 3 rules for inferring interval relationships to aid such questions:

#Infering before and after temporal relationships (between instants and intervals alike)
{?a a ptrec:TemporalData;
    ptrec:instantDT ?timeA. 
 ?b a ptrec:TemporalData;
    ptrec:instantDT ?timeB. ?timeA str:greaterThan ?timeB} 

         => {?a time:intAfter ?b.?b time:intBefore ?a}

{?a a ptrec:TemporalData;
    ptrec:startDT ?startTimeA;
    ptrec:stopDT ?stopTimeA.  
 ?b a ptrec:TemporalData;
    ptrec:startDT ?startTimeB;
    ptrec:stopDT ?stopTimeB. ?startTimeA str:greaterThan ?stopTimeB} 

         => {?a time:intAfter ?b.?b time:intBefore ?a}

#Infering during and contains temporal relationships (between proper intervals)
#Since there is no str:greaterThanOrEqual CWM function, the various permutations
#Are spelled out explicitely
{?a a ptrec:TemporalData;
    ptrec:startDT ?startTimeA;
    ptrec:stopDT ?stopTimeA.  
 ?b a ptrec:TemporalData;
    ptrec:startDT ?startTimeB;
    ptrec:stopDT ?stopTimeB.
 ?startTimeA str:lessThan ?startTimeB. ?stopTimeA str:greaterThan ?stopTimeB} 

         => {?a time:intContains ?b.?b time:intDuring ?a}

{?a a ptrec:TemporalData;
    ptrec:startDT ?startTimeA;
    ptrec:stopDT ?stopTimeA.  
 ?b a ptrec:TemporalData;
    ptrec:startDT ?startTimeB;
    ptrec:stopDT ?stopTimeB.
 ?startTimeA str:equalIgnoringCase ?startTimeB. ?stopTimeA str:greaterThan ?stopTimeB} 

     => {?a time:intContains ?b.?b time:intDuring ?a}

{?a a ptrec:TemporalData;
    ptrec:startDT ?startTimeA;
    ptrec:stopDT ?stopTimeA.  
 ?b a ptrec:TemporalData;
    ptrec:startDT ?startTimeB;
    ptrec:stopDT ?stopTimeB.
 ?startTimeA str:lessThan ?startTimeB. ?stopTimeA str:equalIgnoringCase ?stopTimeB} 

     => {?a time:intContains ?b.?b time:intDuring ?a}

Notice the value in xs:dateTime values being ordered temporally and as unicode, simultaneously. This allows us rely on str:lessThan and str:greaterThan for determining interval intersection and overlap.

Terms such as 'preoperative' (which refer to events that occurred before a specific procedure / operation) and 'postoperative' (events that occurred after a specific procedure / operation), which are core to general medical research nomenclature, can be tied directly into this logic:

{?a a ptrec:TemporalData.  ?b a ptrec:Operation. ?a time:intBefore ?b}
   => {?a ptrec:preOperativeWRT ?b}

{?a a ptrec:TemporalData.  ?b a ptrec:Operation. ?a time:intAfter ?b}
   => {?a ptrec:postOperativeWRT ?b}

Here we introduce two terms (ptrec:preOperativeWRT and ptrec:postOperativeWRT) which relate temporal data with an operation in the same patient record. Using interval relationships as a foundation you can link in domain-specific, temporal vocabulary into your temporal reasoning model, and rely on a reasoner to setup a framework for temporal reasoning.

Imagine the value in using a backward-chaining prover (such as Euler) to logically demonstrate exactly why a specific medication (associated with the date when it was administered) is considered to be preoperative with respect to a qualifying procedure. This would complement the statistical analysis of a case study quite nicely with formal logical proof.

Now, it's worth noting that such a framework (as it currently stands) doesn't allow precision of interval relationships beyond simple intersection and overlap. For instance, in most cases you would be interested primarily in medication administered within a specific length of time. This doesn't really impact the above framework since it is no more than a functional requirement to be able to perform calendar math. Imagine if the built-in properties of CWM were expanded to include functions for performing date math. for instance:

With such a function we can expand our logical framework to include more explicit temporal relationships.
For example, if we only wanted to consider medications that were done 30 days prior to an operation to be considered 'preoperative':

{?a a ptrec:TemporalData;
    ptrec:startDT ?startTimeA;
    ptrec:stopDT ?stopTimeA.  
 ?b a ptrec:Operation;
    ptrec:startDT ?opStartTime;
    ptrec:stopDT ?opStopTime.  
 ?a time:intBefore ?b.
 (?stopTime "-P30D") time:addDT ?preOpMin. ?stopTimeA str:lessThan ?preOpMin}
    => {?a ptrec:preOperativeWRT ?b}

It's worth noting that such an addition (to facilitate calendar math) would be quite useful as a general extension for RDF processors.

For the most part, I think a majority of the requirements needed for temporal reasoning (in any domain) can be accommodated by explicit modeling, because FOPL (the foundation upon which RDF is built) was designed to be expressive enough to represent all human concepts.

Chimezie Ogbuji

via Copia