Thread: [Eulergui-commits] SF.net SVN: eulergui:[1335] trunk/eulergui/html/n3_rules_good_practices. html

SourceForge Headquarters 225 Broadway Suite 1600 San Diego, CA 92101 +1 (858) 422-6466

Revision: 1335
          http://eulergui.svn.sourceforge.net/eulergui/?rev=1335&view=rev
Author:   jmvanel
Date:     2010-03-21 12:37:18 +0000 (Sun, 21 Mar 2010)

Log Message:
-----------
n3_rules_good_practices.html : WIP

Modified Paths:
--------------
    trunk/eulergui/html/n3_rules_good_practices.html

Modified: trunk/eulergui/html/n3_rules_good_practices.html
===================================================================

--- trunk/eulergui/html/n3_rules_good_practices.html	2010-03-19 20:34:52 UTC (rev 1334)
+++ trunk/eulergui/html/n3_rules_good_practices.html	2010-03-21 12:37:18 UTC (rev 1335)
@@ -5,13 +5,329 @@
 <head>
   <meta http-equiv="content-type" content="text/html; charset=iso-8859-1" />
   <title>N3 rules good practices</title>
+  <style type="text/css">
+
+ .CREATION {color:#09b409;
+font-weight:bold; }</style>
   <meta name="generator" content="Amaya, see http://www.w3.org/Amaya/" />
 </head>
 
 <body>
+<p style="font-size: 20pt">N3 rules: good practices, design patterns,
+refactoring</p>
 
-<h2>N3 refactoring: extract part of antecedent</h2>
+<p>\xA9 Jean-Marc Vanel - $Date$</p>
 
+<p>TOC</p>
+
+<div class="toc">
+<ul>
+  <li><a href="#L470">Introduction</a></li>
+  <li><a href="#L587">The limitations of RDF</a>
+    <ul>
+      <li><a href="#L930">Open world assumption</a></li>
+      <li><a href="#L1262">Monotonicity</a></li>
+    </ul>
+  </li>
+  <li><a href="#L593">The limitations of N3 logic</a></li>
+  <li><a href="#L944">Using OWL ontologies</a></li>
+  <li><a href="#L387">Variables patterns</a>
+    <ul>
+      <li><a href="#L398">Quantification</a></li>
+      <li><a href="#L420">Multiple occurences</a></li>
+    </ul>
+  </li>
+  <li><a href="#L180">Design patterns</a>
+    <ul>
+      <li><a href="#L432">Annotate the existing objects</a></li>
+      <li><a href="#L445">Create new objects</a></li>
+    </ul>
+  </li>
+  <li><a href="#L557">Data structures</a></li>
+  <li><a href="#L255">Lists variants</a>
+    <ul>
+      <li><a href="#L215">Lists OWL style</a></li>
+      <li><a href="#L303">Cloning a list</a></li>
+      <li><a href="#L170">Accumulating by recursion</a></li>
+    </ul>
+  </li>
+  <li><a href="#L26">N3 refactorings</a>
+    <ul>
+      <li><a href="#L135">Extract part of antecedent</a></li>
+    </ul>
+  </li>
+</ul>
+</div>
+
+<h2 id="L470">Introduction</h2>
+
+<p>This is a tutorial on writing N3 rules, and various design patterns.</p>
+
+<p>It will address issues of business rules, transformations, validation,
+processing lists, etc..</p>
+
+<p>There is little material directly on the N3 rules, but there are things
+about other rule languages: SWRL, Drools, etc..</p>
+
+<p><a href="http://www.google.fr/search?q=SWRL+design+pattern">Google search
+SWRL+design+pattern</a></p>
+
+<p><a href="http://protege.cim3.net/cgi-bin/wiki.pl?SWRLLanguageFAQ">Prot\xE9g\xE9
+wiki SWRL Language FAQ</a></p>
+
+<p></p>
+
+<p>http://en.wikipedia.org/wiki/Graph_rewriting</p>
+
+<p>http://rewriting.loria.fr/</p>
+
+<p>http://en.wikipedia.org/wiki/Rewriting</p>
+
+<p></p>
+
+<p>Throughout this document, we will put variables corresponding to resource
+(object) creation, that is existentially qualified variables, in <span
+class="CREATION">bold green</span>.</p>
+
+<p><strong>Abreviations</strong></p>
+
+<p>KB == Knowledge Base</p>
+
+<p>LSH == Left Hand Side (a.k.a. antecedent) </p>
+
+<p>RSH == Right Hand Side (a.k.a. consequent) </p>
+
+<p></p>
+
+<h2 id="L587">The limitations of RDF</h2>
+
+<p>The whole point of this paragraph is that there is really no limitation.</p>
+
+<p>First let's say that in terms of expressive power for an object graph data
+language, there is no limitation. Everything that can be expressed in XML, XMI,
+SQL, JSON, can also be expressed in RDF and N3. Moreover, RDF is by design
+adapted to Internet: URI are globally unique identifiers that can be
+dererenced, and it has SWRL for query language and protocol.</p>
+
+<p>Then the principles of the Semantic Web imply the Open World Assumption,
+which is a necessary limitation for the huge and ever changing Web.</p>
+
+<h3 id="L930"><a
+href="http://www.betaversion.org/~stefano/linotype/news/91/">Open world
+assumption</a></h3>
+
+<p>The closed world assumption implies that everything we don&rsquo;t know is
+false, while the open world assumption states that everything we don&rsquo;t
+know is undefined. </p>
+
+<p>The closed world is the point of vue of all databases, except RDF databases.
+Related to closed world assumption, there is a the unique name assumption,
+which mean that a concept has a unique string representation (URI in RDF case);
+or converselly two different string representations are necessarily distinct
+concepts.</p>
+
+<p>The semantic web, N3 logic, and OWL, are based on Open World Assumption. But
+we will see later that there are extension builtins predicates that allow some
+features of closed world.</p>
+
+<h3 id="L1262">Monotonicity</h3>
+
+<p>Monotonicity means that learning a new piece of (non contradictory)
+knowledge cannot reduce the set of what is known ( taken from <a
+href="http://en.wikipedia.org/wiki/Non-monotonic_logic">wikipedia.org /
+Non-monotonic_logic</a> ). This is closely related to Open World Assumption.</p>
+
+<p>A monotonic logic forbids such things as default reasoning ( birds fly
+except ostriches ).</p>
+
+<p>For the same reason, global negation is forbiden ( "no Papuan pilots an
+airplane" ). The authors of N3 logic invented "scoped negation" as a weaker
+form of negation; see the paragraph "Motivation" of <a
+href="http://www.w3.org/DesignIssues/N3Logic">DesignIssues / N3Logic</a> .</p>
+
+<h2 id="L593">The limitations of N3 logic</h2>
+
+<p>The original N3 logic (like SWRL) follows strictly the principles of the
+Semantic Web: Open World Assumption, no Unique Name Assumption, and
+monotonicity.</p>
+
+<p>However, the orignal CWM, Euler and Drools/N3 engines have builtins
+predicates that allow some features of closed world:</p>
+<ul>
+  <li><code>log:equalTo</code> and <code>log:notEqualTo</code> allows a form of
+    Unique Name Assumption</li>
+  <li><code>e:findall</code> allows to by-pass monotonicity; example: find all
+    married couples with more than 1 child: <a
+    href="../test/findall2.n3">test/findall2.n3</a></li>
+</ul>
+
+<p>Breaking monotonicity is not a problem if one is aware of the underlying
+principles. In this respect it is not recommanded to add to the current KB the
+consequences of non-monotonic rules. EulerGUI should flag rules and documents
+containing such rules.</p>
+
+<h2 id="L944">Using OWL ontologies</h2>
+
+<p>add the revelant rules from Euler project. </p>
+
+<p>???</p>
+
+<h2 id="L387">Variables patterns</h2>
+
+<h3 id="L398">Quantification</h3>
+
+<p>existential, universal</p>
+
+<p>variable binding</p>
+
+<p>???</p>
+
+<h3 id="L420">Multiple occurences</h3>
+
+<p>Multiple occurences in LHS means adding constraints, which diminishes the
+set of matching statements for the rule.</p>
+
+<p>Multiple occurences in RHS means adding new statements to the KB.</p>
+
+<h2 id="L180">Design patterns</h2>
+
+<h3 id="L432">Annotate the existing objects</h3>
+
+<p>This is the simplest pattern, and the most efficient on all engines. It is
+particuliar adapted to problems with a 1 to 1 correspondence between the
+asserted objects and the inferred objects. It is opposed to the following
+pattern "Create new objects".</p>
+
+<h3 id="L445">Create new objects</h3>
+
+<p>keep track</p>
+
+<p>often not possible with Euler engine</p>
+
+<p>???</p>
+
+<h2 id="L557">Data structures</h2>
+
+<p>No sets ? Why ? Because ordinary RDF properties act as sets.</p>
+
+<p>RDF is a bunch of associative arrays (called maps in Java and C++). ????</p>
+
+<h2 id="L255">Lists variants</h2>
+
+<p>RDF lists, OWL, linked list.</p>
+
+<p>transitive properties</p>
+
+<h3 id="L215">Lists OWL style</h3>
+
+<p>????</p>
+
+<p>There is a paper describing the list pattern in OWL:</p>
+
+<p>Drummond, N., Rector, A.L., Stevens, R., Moulton, G., Horridge, M., Wang, H,
+Seidenberg, J. Putting OWL in Order: Patterns for Sequences in OWL, in 2nd OWL
+Experiences and Directions Workshop, Athens, GA</p>
+
+<p><a
+href="http://www.webont.org/owled/2006/acceptedLong/submission_12.pdf">http://www.webont.org/owled/2006/acceptedLong/submission_12.pdf</a></p>
+
+<p></p>
+
+<h3 id="L303">Cloning a list</h3>
+
+<p>Here is an example with user specific :first and :rest properties , <a
+href="../examples/clone-list-forward.n3">clone-list-forward.n3</a>. It clones
+the list in the forward direction.</p>
+<pre>{ ?L :first ?F .
+  ?L :rest  ?R .
+  ?LC :clone_of ?L .
+} =&gt; {
+  ?LC :first ?F .
+  ?LC :rest <strong><span class="CREATION">?<span style="font-weight:bold;"></span>RC</span></strong> .
+  <span class="CREATION">?RC</span> :clone_of ?R .
+} .</pre>
+
+<p>It must be provided with a start statement, e.g.:</p>
+<pre>:newNode :clone_of :existingNode .</pre>
+
+<p></p>
+
+<p>Due to limitations in Euler/Eye engine, it is not possible to generate
+rdf:first and rdf:rest properties. However, this is possible </p>
+
+<p>with Drools/N3 engine. With both engines, it is possible to to clone an RDF
+list using a rule similar to the above, matching rdf:first and rdf:rest
+properties in the LSH : <a
+href="../examples/clone-rdf-list-forward.n3">clone-rdf-list-forward.n3</a> .</p>
+
+<p></p>
+
+<p>Here is another example with user specific :first and :rest properties, <a
+href="../examples/clone-list.n3">clone-list.n3</a>. It clones the list in the
+"backward" direction, that is starting from the end.</p>
+<pre>{ # initial condition
+  ?L :first ?F .
+  ?L :rest  :nil .
+} =&gt; {
+  <span class="CREATION">?LC</span> :first ?F .
+  <span class="CREATION">?LC</span> :rest_clone :nil .
+  <span class="CREATION">?LC</span> :clone_of ?L .
+} .
+{ # recursion
+  ?L :first ?F .
+  ?L :rest  ?R .
+  ?RC :clone_of ?R .
+} =&gt; {
+  ?LC :first ?F .
+  ?LC :rest_clone  ?RC .
+  ?LC :clone_of ?L .
+} .</pre>
+
+<p></p>
+
+<p>Note the difference in semantics with the preceding example, where the
+cloning was triggered by a "start statement". Here, whenever there is a list
+made with the :first and :rest properties, a clone is created.</p>
+
+<p></p>
+
+<p>These examples are just design patterns. I real life, for cloning to be
+useful, domain logic will be added, like:</p>
+<ul>
+  <li>filtering some items in the list (e.g. creating a list of good customers
+    from a list of customers)</li>
+  <li>assigning some properties to the cloned nodes (e.g. associating a sales
+    representative to each customer in the list)</li>
+</ul>
+
+<p></p>
+
+<h3 id="L170">Accumulating by recursion</h3>
+
+<p>Itinerary is a linked list, and each Itinerary is also a step.</p>
+<pre>{
+  ?I a itin:Itinerary .
+  ?I itin:distance ?D .
+  ?I list:hasNext ?N .
+  ?N :totalDistance ?NT .
+  (?D ?NT) math:sum ?T .
+} =&gt; {
+  ?I :totalDistance ?T .
+} .
+
+{
+  ?I a itin:Itinerary .
+  ?I list:hasNext itin:emptyItinerary .
+} =&gt; {
+  ?I :totalDistance 0.0 .
+} .</pre>
+
+<p></p>
+
+<h2 id="L26">N3 refactorings</h2>
+
+<h3 id="L135">Extract part of antecedent</h3>
+
 <p><strong>Motivation</strong></p>
 
 <p>Suppose the original rule OR is writen as :</p>


This was sent by the SourceForge.net collaborative development platform, the world's largest Open Source development site.




Thread: [Eulergui-commits] SF.net SVN: eulergui:[1335] trunk/eulergui/html/n3_rules_good_practices. html

eulergui-commits