<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
  <head>
    <meta content="text/html; charset=ISO-8859-1"
      http-equiv="Content-Type">
  </head>
  <body bgcolor="#ffffff" text="#000000">
    <br>
    I am not sure I understand what you mean by "random order" but I
    guess it has to do with my ArrayList result type.<br>
    What I had in mind is to put all sentences in a TreeSet during the
    "action" method, and finally issue an ArrayList result object by
    iterating over the TreeSet and grouping sentences.<br>
    My first guess was that such an accumulator might be faster than a
    construction of windows using rules.<br>
    However I admit your suggestion is very elegant, and I thank you for
    that! I am probably still too imperative-minded...<br>
    <br>
    Regards,<br>
    <br>
    Bruno.<br>
    <br>
    Le 19/08/2011 16:05, Wolfgang Laun a &eacute;crit&nbsp;:
    <blockquote
cite="mid:CANaj1LfZGMP6O7hMZdAKrTY3V1XxetTmTPK83DqjA=_Jj-FEVQ@mail.gmail.com"
      type="cite">How would you write "buildwindows", given that its
      "action" method would be called once for each Sentence, in random
      order?<br>
      <br>
      It's very simple to write a very small set of rules to construct
      all SentenceWindow facts of size 1 and then to extend them to any
      desired size, depending on some parameter.<br>
      1. Given a Sentence and no Window beginning with it, create a
      Window of length 1.<br>
      2. Given a Window of size n &lt; desiredSize and given a Sentence
      immediately following it, extend the Window to one of size n+1.<br>
      3a. For any Window of desiredSize, inspect it for "closely
      situated ManualAnnotations".<br>
      3b. If ManualAnnotations have been associated with their
      containing Sentences up-front, you just need to find Windows with
      more than 1 ManualAnnotation, adding them in the RHS of rule 2
      above.<br>
      <br>
      -W<br>
      <br>
      <br>
      <div class="gmail_quote">
        2011/8/19 Bruno Freudensprung <span dir="ltr">&lt;<a
            moz-do-not-send="true"
            href="mailto:bruno.freudensprung@temis.com">bruno.freudensprung@temis.com</a>&gt;</span><br>
        <blockquote class="gmail_quote" style="border-left: 1px solid
          rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left:
          1ex;">
          <div bgcolor="#ffffff" text="#000000"> <br>
            Hi Wolfgang,<br>
            <br>
            Thanks for your answer.<br>
            Sentences are not contiguous (might be some space characters
            in between) but manual annotations cannot overlap sentences
            (interpret "overlap" in terms of Drools Fusion terminology).<br>
            If I had an "inside" operator, do you think the following
            accumulate option could be better? <br>
            <br>
            <tt>when<br>
              &nbsp;&nbsp;&nbsp; <b>$result : ArrayList() from accumulate ( $s:
                Sentence(), buildwindows($s))</b><br>
            </tt><tt>&nbsp;&nbsp;&nbsp;<b> $w : SentenceWindows () </b></tt><b><tt>from

                $result</tt></b><br>
            <tt> </tt><tt>&nbsp;&nbsp;&nbsp; a1 : ManualAnnotation (this <b>inside </b>$w)<br>
              &nbsp;&nbsp;&nbsp; a2 : ManualAnnotation (this != a1, </tt><tt>this <b>inside
              </b>$w</tt><tt>)<br>
            </tt><tt>then<br>
              &nbsp;&nbsp;&nbsp; ... </tt><tt>do something with a1 and a2 since they
              are "close" to each other</tt><br>
            <tt> end</tt><br>
            <br>
            Does anyone know something about accumulator parametrization
            (looking at the source code it does not seem to be possible,
            though)?<br>
            Maybe a syntax inspired of operator parametrization could be
            nice:<br>
            <br>
            <tt> &nbsp;&nbsp;&nbsp; $result : ArrayList() from accumulate ( $s:
              Sentence(), <b> buildwindows[3]($s)</b>)<br>
            </tt><br>
            Best regards,<br>
            <br>
            Bruno.<br>
            <br>
            Le 19/08/2011 13:55, Wolfgang Laun a &eacute;crit&nbsp;:
            <blockquote type="cite">There are some details that one
              should consider before deciding on a particular
              implementation technique.<br>
              <ul>
                <li>Are all Sentences contiguous, i.e., s1.end = pred(
                  s2.start )</li>
                <li>Can a ManualAnnotation start on one Sentence and end
                  in the next or any further successor?</li>
              </ul>
              As in all problems where constraints depend on an order
              between facts, performance is going to be a problem with
              increasing numbers of Sentences and ManualAnnotations.<br>
              <br>
              Your accumulate plan could be a very inefficient approach.
              Creating O(N*N) pairs and then looking for an overlapping
              window is much worse than looking at each window, for
              instance. But it depends on the expected numbers for both.<br>
              <br>
              -W<br>
              <br>
              <br>
              <br>
              <div class="gmail_quote">2011/8/19 Bruno Freudensprung <span
                  dir="ltr">&lt;<a moz-do-not-send="true"
                    href="mailto:bruno.freudensprung@temis.com"
                    target="_blank">bruno.freudensprung@temis.com</a>&gt;</span><br>
                <blockquote class="gmail_quote" style="border-left: 1px
                  solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex;
                  padding-left: 1ex;">
                  <div bgcolor="#ffffff" text="#000000"> Hello,<br>
                    <br>
                    I am trying to implement rules handling "Sentence",
                    "ManualAnnotation" objects (imagine someone
                    highligthing words of the document). Basically
                    "Sentence" objects have "start" and "end" positions
                    (fields) into the text of a document, and they are
                    Comparable according to their location into the
                    document.<br>
                    <br>
                    I need to write rules using the notion "window of
                    consecutive sentences". <br>
                    <br>
                    Basically I am not very interested by those
                    "SentenceWindow" objects, I just need them to define
                    a kind of proximity between "ManualAnnotation"
                    objects.<br>
                    What I eventually need in the "when" of my rule is
                    something like:<br>
                    <br>
                    <tt>when<br>
                      &nbsp;&nbsp;&nbsp; ... maybe something creating the windows<br>
                      &nbsp;&nbsp;&nbsp; a1 : ManualAnnotation ()<br>
                      &nbsp;&nbsp;&nbsp; a2 : ManualAnnotation (this != a1)<br>
                      &nbsp;&nbsp;&nbsp; SentenceWindow (this includes a1, this
                      includes a2)<br>
                      then<br>
                      &nbsp;&nbsp;&nbsp; ... do something with a1 and a2 since they are
                      "close" to each other<br>
                      end</tt><br>
                    <br>
                    As I don't know the "internals" of Drools, I would
                    like to have your opinion about what the best
                    "idiom":<br>
                    <ul>
                      <li>create all SentenceWindow objects and insert
                        them in the working memory, then write rules
                        against all the facts (SentenceWindow and
                        ManualAnnotation)<br>
                      </li>
                      <li>implement an accumulator that will create a
                        list of&nbsp; SentenceWindow object </li>
                    </ul>
                    <br>
                    The first option could look like:<br>
                    <br>
                    <code></code><code>rule "Create sentence windows"<br>
                      &nbsp; &nbsp;when<br>
                      &nbsp;&nbsp;&nbsp; &nbsp; # find 3 consecutive sentences<br>
                      &nbsp; &nbsp;&nbsp; &nbsp;s1 : Sentence()<br>
                      &nbsp; &nbsp;&nbsp; &nbsp;s2 : Sentence(this &gt; s1)<br>
                      &nbsp; &nbsp;&nbsp; &nbsp;s3 : Sentence(this &gt; s2)<br>
                      &nbsp; &nbsp;&nbsp; &nbsp;not Sentence(this != s2 &amp;&amp; &gt; s1
                      &amp;&amp; &lt; s3)<br>
                      &nbsp; &nbsp;then<br>
                      &nbsp; &nbsp;&nbsp; &nbsp;SentenceWindow swindow = new
                      SentenceWindow();<br>
                      &nbsp; &nbsp;&nbsp; &nbsp;swindow.setStart(s1.getStart());<br>
                      &nbsp; &nbsp;&nbsp; &nbsp;swindow.setTheend(s3.getEnd());<br>
                      &nbsp; &nbsp;&nbsp; &nbsp;insert(swindow);<br>
                      end</code><br>
                    <br>
                    ... Then use the first rule "as is".<br>
                    <br>
                    The accumulator option could look like (I am not
                    really sure the syntax is correct) :<br>
                    <br>
                    <tt>when<br>
                      &nbsp;&nbsp;&nbsp; <b>$result : ArrayList() from accumulate (
                        $s: Sentence(), buildwindows($s))</b><br>
                    </tt><tt>&nbsp;&nbsp;&nbsp; a1 : ManualAnnotation ()<br>
                      &nbsp;&nbsp;&nbsp; a2 : ManualAnnotation (this != a1)<br>
                      &nbsp;&nbsp;&nbsp;<b> SentenceWindows (this includes a1, this
                        includes a2) </b></tt><b><tt>from $result</tt></b><br>
                    <tt> then<br>
                      &nbsp;&nbsp;&nbsp; ... </tt><tt>do something with a1 and a2
                      since they are "close" to each other</tt><br>
                    <tt> end</tt><br>
                    <br>
                    Is it possible to decide if one way is best than the
                    other?<br>
                    <br>
                    And one last question: it is possible to
                    "parametrize" an accumulator (in order to provide
                    the number of sentences that should be put in the
                    windows)?<br>
                    I mean something like:<br>
                    <br>
                    <tt>when<br>
                      &nbsp;&nbsp;&nbsp; $result : ArrayList() from accumulate ( $s:
                      Sentence(), <b>buildwindows(3,</b> $s))<br>
                    </tt><br>
                    <br>
                    Thanks in advance for you insights,<br>
                    <br>
                    Best regards,<br>
                    <font color="#888888"> <br>
                      Bruno.<br>
                    </font></div>
                  <br>
                  _______________________________________________<br>
                  rules-users mailing list<br>
                  <a moz-do-not-send="true"
                    href="mailto:rules-users@lists.jboss.org"
                    target="_blank">rules-users@lists.jboss.org</a><br>
                  <a moz-do-not-send="true"
                    href="https://lists.jboss.org/mailman/listinfo/rules-users"
                    target="_blank">https://lists.jboss.org/mailman/listinfo/rules-users</a><br>
                  <br>
                </blockquote>
              </div>
              <br>
              <pre><fieldset></fieldset>
_______________________________________________
rules-users mailing list
<a moz-do-not-send="true" href="mailto:rules-users@lists.jboss.org" target="_blank">rules-users@lists.jboss.org</a>
<a moz-do-not-send="true" href="https://lists.jboss.org/mailman/listinfo/rules-users" target="_blank">https://lists.jboss.org/mailman/listinfo/rules-users</a>
</pre>
            </blockquote>
            <br>
          </div>
          <br>
          _______________________________________________<br>
          rules-users mailing list<br>
          <a moz-do-not-send="true"
            href="mailto:rules-users@lists.jboss.org">rules-users@lists.jboss.org</a><br>
          <a moz-do-not-send="true"
            href="https://lists.jboss.org/mailman/listinfo/rules-users"
            target="_blank">https://lists.jboss.org/mailman/listinfo/rules-users</a><br>
          <br>
        </blockquote>
      </div>
      <br>
      <pre wrap="">
<fieldset class="mimeAttachmentHeader"></fieldset>
_______________________________________________
rules-users mailing list
<a class="moz-txt-link-abbreviated" href="mailto:rules-users@lists.jboss.org">rules-users@lists.jboss.org</a>
<a class="moz-txt-link-freetext" href="https://lists.jboss.org/mailman/listinfo/rules-users">https://lists.jboss.org/mailman/listinfo/rules-users</a>
</pre>
    </blockquote>
    <br>
  </body>
</html>