<div dir="ltr"><br><br><div class="gmail_quote"><div dir="ltr">Am Fr., 21. Sep. 2018 um 23:12 Uhr schrieb <a href="mailto:goodman.m.w@gmail.com">goodman.m.w@gmail.com</a> &lt;<a href="mailto:goodman.m.w@gmail.com">goodman.m.w@gmail.com</a>&gt;:<br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr"><div dir="ltr"><div dir="ltr"><div style="font-family:arial,helvetica,sans-serif">Returning to this thread so it doesn&#39;t fade away...<br></div><div style="font-family:arial,helvetica,sans-serif"><br></div><div style="font-family:arial,helvetica,sans-serif">I don&#39;t understand the complexities well enough to have an opinion on whether we should use YADU, Guy&#39;s variant, or something else.</div></div></div></div></blockquote><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr"><div dir="ltr"><div dir="ltr"><div style="font-family:arial,helvetica,sans-serif"> I&#39;m concerned, however, that if it is sufficiently complicated to work out the ramifications of the use of defaults, it would outweigh the benefits from the perspective of the grammar developer. With this in mind I would (...expect that a grammar developer would...) prefer something more predictable over something slightly more powerful. I&#39;d even be happy with something more restrictive than YADU if it were still useful, easy to understand, and could give informative error messages at compile time. That is, can we somehow proscribe the complicated corner cases with a more restrictive definition?<br></div></div></div></div></blockquote><div><br></div><div>
I think it would make sense to have a number of use cases (or unit 
tests) to see whether any variant of defaults would help the developer. 
 I proposed my variant to deal with the use case Emily gave -- but there
 could well be other cases I haven&#39;t thought of, as well as side effects
 I haven&#39;t thought of!  I&#39;m not sure that a more restrictive version of YADU could deal with Emily&#39;s use case.<br></div><div><br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr"><div dir="ltr"><div dir="ltr"><div style="font-family:arial,helvetica,sans-serif"></div><div style="font-family:arial,helvetica,sans-serif"><br></div><div style="font-family:arial,helvetica,sans-serif">Regarding the syntax of these things, Emily and Guy use /#coref, whereas Copestake 2002, the ParisDefeasibleConstraints wiki, and the LKB use /l (lowercase L for &quot;lexical&quot;, which is specified by the *description-persistence* parameter in the LKB). I don&#39;t really understand the difference, but it seems the latter doesn&#39;t require a paired coreference, as shown in regverb in the standard example below:<br></div><div style="font-family:arial,helvetica,sans-serif"><br></div><div style="font-family:arial,helvetica,sans-serif">    verb := *top* &amp;<br>    [ PAST /l #pp,<br>      PASTP #p /l #pp,<br>      PASSP #p /l #pp ].<br><br>    regverb := verb &amp;<br>    [ PAST /l &quot;ed&quot; ] .</div><div style="font-family:arial,helvetica,sans-serif"><br></div><div style="font-family:arial,helvetica,sans-serif">If these are in fact different instances of the same mechanism, then maybe &quot;/#coref&quot; is just &quot;/ #coref&quot; (i.e., a persistent-default operator followed by an actual coreference)?<br></div></div></div></div></blockquote><div><br></div><div>The YADU paper (Lascarides and Copestake, 1999) also uses just &quot;/&quot; rather than &quot;/l&quot; -- although that paper gives typeset AVMs, rather than machine-readable tdl.  I&#39;d be happy to stick to &quot;/l&quot; if that would reduce confusion!<br></div><div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr"><div dir="ltr"><div dir="ltr"><div style="font-family:arial,helvetica,sans-serif"></div><div style="font-family:arial,helvetica,sans-serif"><br></div><div style="font-family:arial,helvetica,sans-serif">On a related note, I recently added to PyDelphin the ability to serialize TDL, so I can imagine writing code that would transform a grammar with defaults to one without as some kind of compile-time preprocessor. It would be great if we could agree on and provide implementations of defaults before the next summit.<br></div></div></div></div></blockquote><div><br></div><div>Yes!  So far I haven&#39;t been able to find explicit documentation of how defaults are handled.  My current understanding is that, the compile-time processor needs to expand type and instance definitions to include all inherited constraints -- and then the (possibly inherited) default constraints on instances are turned into hard constraints, while the (possibly inherited) default constraints on types are discarded.  I would be happy to be corrected here if I&#39;m wrong!<br></div><div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr"><div dir="ltr"><div dir="ltr"><div style="font-family:arial,helvetica,sans-serif"></div></div></div></div><br><div class="gmail_quote"><div dir="ltr">On Thu, Sep 6, 2018 at 4:17 AM Guy Emerson &lt;<a href="mailto:gete2@cam.ac.uk" target="_blank">gete2@cam.ac.uk</a>&gt; wrote:<br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr"><div>What I&#39;ve proposed isn&#39;t credulous default unification, although it does use it as part of its definition, since DefFS does.  The YADU paper does describe credulous default unification (definition 9, on page 69, or page 15 of the pdf), citing Carpenter.<br></div><div><br></div><div>Credulously unifying [ F #1, G #1 ] into [ F.P c, G.P d ] just yields the singleton set without any new constraints, { 
[ F.P c, G.P d ] }, because the only piece of information in 
[ F #1, G #1 ] that is not already in 
[ F.P c, G.P d ] is the re-entrancy, but this re-entrancy is incompatible.  It&#39;s crucial that 
[ F #1, G #1 ] does not contain the paths F.Q and G.Q, or else these *would* be incorporated in credulous unification -- but this follows Emily&#39;s example with valence-min, where the relevant feature paths are similarly not there.  This is why I proposed first expanding 
[ F #1, G #1 ] by credulously unifying in the opposite direction.<br></div><div><br><div class="gmail_quote"><div dir="ltr">Am Mi., 5. Sep. 2018 um 18:05 Uhr schrieb Ann Copestake &lt;<a href="mailto:aac10@cl.cam.ac.uk" target="_blank">aac10@cl.cam.ac.uk</a>&gt;:<br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
  
    
  
  <div text="#000000" bgcolor="#FFFFFF">
    <p>this seems like it might be credulous default unification?  can&#39;t
      remember whether we described that in the YADU paper or not - I
      think the idea comes from Bob Carpenter.  Sorry I am just
      procrastinating and don&#39;t have time to read/think properly, or
      check.</p>
    <p>Cheers,</p>
    <p>Ann<br>
    </p>
    <br>
    <div class="m_-8593124814449095464m_4777295698704568155m_-4850224290236657958moz-cite-prefix">On 05/09/18 17:56, Guy Emerson wrote:<br>
    </div>
    <blockquote type="cite">
      
      <div dir="ltr">
        <div dir="ltr">
          <div dir="ltr">
            <div dir="ltr">
              <div dir="ltr">
                <div dir="ltr">
                  <div dir="ltr">
                    <div dir="ltr">
                      <div dir="ltr">
                        <div dir="ltr">
                          <div dir="ltr">I&#39;m going to adapt Alex&#39;s
                            extended Nixon diamond example from <a href="http://moin.delph-in.net/StanfordDefaults" target="_blank">http://moin.delph-in.net/StanfordDefaults</a>
                            , because it&#39;s a minimal working example
                            that captures the essence of Emily&#39;s
                            example.</div>
                          <div dir="ltr"><br>
                          </div>
                          <div>We have:</div>
                          <div><br>
                          </div>
                          <div>a := *top* &amp;</div>
                          <div>  [ F /#1,</div>
                          <div>    G /#1 ].</div>
                          <div><br>
                          </div>
                          <div>x := *top* &amp;</div>
                          <div>  [ P *top*,</div>
                          <div>    Q *top* ].</div>
                          <div><br>
                          </div>
                          <div>c := *top*.</div>
                          <div>d := *top*.<br>
                          </div>
                          <div><br>
                          </div>
                          <div>b := a &amp;</div>
                          <div>  [ F.P c,</div>
                          <div>    G.P d ].</div>
                          <div><br>
                          </div>
                          <div>I take it that the behaviour we would
                            like to have is that the default re-entrancy
                            between F and G is pushed down (in b) to a
                            re-entrancy between F.Q and G.Q.  As I
                            understand, YADU ( <a href="http://www.aclweb.org/anthology/J99-1002" target="_blank">http://www.aclweb.org/anthology/J99-1002</a>
                            ) would discard the re-entrancy, because
                            it&#39;s not compatible with the hard
                            constraints.  (Also, note that I have
                            deliberately left the features P and Q
                            outside of the definition of a, to mimic
                            valence-min in Emily&#39;s example -- we can&#39;t
                            represent the re-entrancy between F.Q and
                            G.Q in a, but only in b.)</div>
                          <div><br>
                          </div>
                          <div>I have a tentative proposal -- and I say
                            tentative, because Ann has already warned
                            that YADU can blow up, and what I&#39;m about to
                            suggest would be probably be even more
                            likely to blow up.  So here be explosive
                            dragons!</div>
                          <div><br>
                          </div>
                          <div>But with that warning aside... YADU
                            represents a defeasible feature structure in
                            two parts - a normal feature structure
                            (containing the hard constraints), and a
                            &quot;tail&quot; of default constraints.  In the above
                            example, only a has a non-empty tail, which
                            just has one thing in it (the re-entrancy). 
                            With YADU, unifying two defeasible
                            structures involves first unifying their
                            hard structures, combining the tails, and
                            then discarding anything in the tail that&#39;s
                            incompatible with the hard structure.  My
                            proposed extension of YADU is not to discard
                            incompatible constraints in the tail, but
                            rather &quot;expand&quot; them and keep whatever
                            compatible constraints are in this expanded
                            set.</div>
                          <div><br>
                          </div>
                          <div>This &quot;expansion&quot; of constraints doesn&#39;t
                            require new algebraic operations, because we
                            can use what&#39;s already in YADU -- in
                            particular, the DefFS operation (Definition
                            12 on page 70, which is page 16 in the
                            pdf).  Essentially, DefFS takes a hard
                            structure and a set of possibly
                            contradictory default constraints, and adds
                            as many of the default constraints as
                            possible without creating a contradiction. 
                            My proposal is to apply DefFS *backwards* --
                            if we unify two hard structures and find
                            that some default constraint in the tail is
                            now incompatible with the unified hard
                            structure, we can use DefFS to &quot;expand&quot; the
                            default constraint.  We treat the default
                            constraint as the hard structure in DefFS,
                            and we treat the unified hard structure as
                            the tail in DefFS.  This tells us how much
                            of the unified hard structure is compatible
                            with the default constraint -- in other
                            words, we have &quot;expanded&quot; the default
                            constraint in the context of the unified
                            hard structure.  We can now decompose this
                            expanded structure into individual
                            constraints, and add them to the new tail,
                            as long as they don&#39;t contradict the new
                            unified hard structure.</div>
                          <div><br>
                          </div>
                          <div>This might be clearest looking at the
                            above example.  We would like to unify b&#39;s
                            feature structure with a&#39;s feature
                            structure.  This gives us the following hard
                            structure, and tail (I am ignoring the types
                            of tail elements, for simplicity, because we
                            aren&#39;t considering when one default
                            overrides another default):<br>
                          </div>
                          <div><br>
                          </div>
                          [ F x &amp; [ P c,<br>
                                    Q *top* ],<br>
                            G x &amp; [ P d,<br>
                                    Q *top* ] ].
                          <div>tail: { F=G }</div>
                          <div><br>
                          </div>
                          <div>To apply DefFS, we first credulously add
                            the hard structure&#39;s constraints to the tail
                            element.  This gives a pair of structures
                            (which differ in the value of F.P):</div>
                          <div><br>
                          </div>
                          <div>[ F #1 &amp; x &amp; [ P c,<br>
                                           Q *top* ],<br>
                              G #1 ].</div>
                          <div><br>
                          </div>
                          <div>[ F #1 &amp; x &amp; [ P d,<br>
                                           Q *top* ],<br>
                              G #1 ].</div>
                          <div><br>
                          </div>
                          <div>DefFS gives us the generalisation of
                            these two structures, which is:</div>
                          <div><br>
                          </div>
                          <div>[ F #1 &amp; x &amp; [ P *top*,<br>
                                           Q *top* ],<br>
                              G #1 ].<br>
                          </div>
                          <div><br>
                          </div>
                          <div>We can break this up into individual
                            constraints: { F:x, G:x, F=G, F.P=G.P,
                            F.Q=G.Q }</div>
                          <div><br>
                          </div>
                          <div>Of these five constraints, the first two
                            are already part of the unified hard
                            structure, and the second two are
                            incompatible with it.  So we are left with a
                            single element to keep in the tail:
                            F.Q=G.Q.  So, expanding elements in the tail
                            rather than discarding them gives:</div>
                          <div><br>
                          </div>
                          <div>[ F x &amp; [ P c,<br>
                                      Q *top* ],<br>
                              G x &amp; [ P d,<br>
                                      Q *top* ] ].
                            <div>tail: { F.Q=G.Q }</div>
                          </div>
                          <div><br>
                          </div>
                          <div>Finally, the last step (as with YADU) is
                            to apply DefFS, which would give the
                            following compiled structure used at runtime
                            (supposing I&#39;ve understood what is supposed
                            to happen at compile time -- I wasn&#39;t sure
                            where to look for this documentation):</div>
                          <div><br>
                          </div>
                          <div>[ F x &amp; [ P c,<br>
                                      Q #1 ],<br>
                              G x &amp; [ P d,<br>
                                      Q #1 ] ].<br>
                          </div>
                        </div>
                      </div>
                    </div>
                  </div>
                </div>
              </div>
            </div>
          </div>
        </div>
      </div>
      <br>
      <div class="gmail_quote">
        <div dir="ltr">Am Di., 4. Sep. 2018 um 23:10 Uhr schrieb Emily
          M. Bender &lt;<a href="mailto:ebender@uw.edu" target="_blank">ebender@uw.edu</a>&gt;:<br>
        </div>
        <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
          <div dir="ltr">Dear Mike,
            <div><br>
            </div>
            <div>Thanks for bringing up this issue.  At the 2010
              DELPH-IN Paris Summit, Ann and I had a further
              conversation about this, from which I took the homework of
              typing up what it is I&#39;d like to have (as a grammar
              developer, and especially from the point of view of the
              Matrix) wrt to defeasible constraints.  Here&#39;s what I
              wrote down later that year (Oct 27):</div>
            <div><br>
            </div>
            <div><span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">Dear
                Ann,</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">Here,
                with much more delay than I intended, is the write up</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">I
                promised of my (reconstruction of my) understanding of
                where</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">we
                ended up in our discussion of<span> </span></span><span class="m_-8593124814449095464m_4777295698704568155m_-4850224290236657958m_-2966151228754498155gmail-il" style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">defeasible</span><span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"><span> </span>identity<span> </span></span><span class="m_-8593124814449095464m_4777295698704568155m_-4850224290236657958m_-2966151228754498155gmail-il" style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">constraints</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">over
                crepes in Paris.</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">First,
                why I want it:</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">In
                lexical rules, we want to be able say (like in SWB) that</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">the
                value of certain features (HOOK, CAT, ARG-ST) is shared</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">between
                the mother and the daughter unless the rule contradicts</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">this. 
                If the rule does contradict it, then we want only the
                information</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">specifically
                stated as such to change, and the rest &quot;around&quot; it,</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">to
                be shared.</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">For
                a concrete example, take a hypothetical lexical rule
                that</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">changes
                the case on the first complement from acc to dat.</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">First,
                here&#39;s the general lex rule type:</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">lex-rule
                := phrase-or-lexrule &amp; word-or-lexrule &amp;</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> 
                [ NEEDS-AFFIX bool,</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> 
                  SYNSEM.LOCAL.CONT [ RELS [ LIST #first,</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> 
                                           LAST #last ],</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> 
                                      HCONS [ LIST #hfirst,</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> 
                                              LAST #hlast ] ],</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> 
                  DTR #dtr &amp; word-or-lexrule &amp;</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> 
                      [ SYNSEM.LOCAL.CONT [ RELS [ LIST #first,</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> 
                                                   LAST #middle ],</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> 
                                            HCONS [ LIST #hfirst,</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> 
                                                    LAST #hmiddle ] ],</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> 
                        ALTS #alts ],</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> 
                  C-CONT [ RELS [ LIST #middle,</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> 
                                  LAST #last ],</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> 
                           HCONS [ LIST #hmiddle,</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> 
                                   LAST #hlast ]],</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> 
                  ALTS #alts,</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> 
                  ARGS &lt; #dtr &gt; ].</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">And
                a subtype with the<span> </span></span><span class="m_-8593124814449095464m_4777295698704568155m_-4850224290236657958m_-2966151228754498155gmail-il" style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">defeasible</span><span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"><span> </span>identity
                indicated</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">(using
                /# for now):</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span class="m_-8593124814449095464m_4777295698704568155m_-4850224290236657958m_-2966151228754498155gmail-il" style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">defeasible</span><span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">-identity-lex-rule
                := lex-rule &amp;</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> 
                [<span> </span></span><a href="http://synsem.local.cat/" rel="noreferrer" style="color:rgb(17,85,204);font-size:12.8px;background-color:rgb(255,255,255)" target="_blank">SYNSEM.LOCAL.CAT</a><span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"><span> </span>/#cat,</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> 
                  ARG-ST /#arg-st,</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> 
                  C-CONT.HOOK /#hook,</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> 
                  DTR [ LOCAL [ CAT /#cat,</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> 
                                         CONT.HOOK /#hook ],</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> 
                           ARG-ST /#arg-st ]].</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">The
                lex rule definition itself would just look like this:</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">acc-to-dat-obj-lex-rule
                := lex-rule &amp;</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> [
                SYNSEM.LOCAL.CAT.COMPS.FIRST.</span><span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">LOCAL.CAT.HEAD.CASE
                dat,</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> 
                 DTR.SYNSEM.LOCAL.CAT.COMPS.</span><span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">FIRST.LOCAL.CAT.HEAD.CASE
                acc ].</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">The
                intended behavior is for that to compile into a rule
                that</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">includes<span> </span></span><span class="m_-8593124814449095464m_4777295698704568155m_-4850224290236657958m_-2966151228754498155gmail-il" style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">constraints</span><span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"><span> </span>like
                these (I&#39;m sure I&#39;m missing some here):</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">acc-to-dat-obj-lex-rule
                (expanded):</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> [<span> </span></span><a href="http://synsem.local.cat/" rel="noreferrer" style="color:rgb(17,85,204);font-size:12.8px;background-color:rgb(255,255,255)" target="_blank">SYNSEM.LOCAL.CAT</a><span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"><span> </span>[
                HEAD #head,</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> 
                                                    VAL [ SPR #spr,</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> 
                                                             SPEC #spec,</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> 
                                                             SUBJ #subj,</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> 
                                                             COMPS [
                REST #rest,</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> 
                                                                       
                   FIRST [</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">NON-LOCAL
                #non-local,</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> 
                LOCAL [ CONT #cont,</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> 
                              CAT [ VAL #val,</span></div>
            <div><span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> 
                                        AGR #agr,</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> 
                                        HEAD.CASE dat ]]]]]],</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> 
                C-CONT.HOOK #hook,</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> 
                ARG-ST #arg-st,</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> 
                DTR [ SYNSEM.LOCAL [ CONT.HOOK #hook,</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> 
                                                     CAT [ HEAD #head,</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> 
                                                               VAL [ SPR
                #spr,</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> 
                                                                       
                SPEC #spec,</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> 
                                                                       
                SUBJ #subj,</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> 
                                                                       
                COMPS [ REST #rest,</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> FIRST
                [ NON-LOCAL #non-local,</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> 
                          LOCAL [ CONT #cont,</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> 
                                        CAT [ VAL #val,</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> 
                                                 AGR #agr,</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> 
                                                 HEAD.CASE acc ]]]]]]],</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"> 
                         ARG-ST #arg-st ]].</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">What
                I remember from Paris is that we decided it would be
                best to</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">encode
                these<span> </span></span><span class="m_-8593124814449095464m_4777295698704568155m_-4850224290236657958m_-2966151228754498155gmail-il" style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">constraints</span><span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"><span> </span>not
                directly in the type definition as</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">I
                did above in<span> </span></span><span class="m_-8593124814449095464m_4777295698704568155m_-4850224290236657958m_-2966151228754498155gmail-il" style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">defeasible</span><span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">-identity-lex-rule
                but in a collateral file that</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">instructs
                the LKB to do something special with certain feature
                paths</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">on
                instances of certain types at compile time.</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">We
                also worked out that we would only be able to &quot;push
                down&quot; the identity</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span class="m_-8593124814449095464m_4777295698704568155m_-4850224290236657958m_-2966151228754498155gmail-il" style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">constraint</span><span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"><span> </span>to
                features that were necessitated by the types invoked</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">in
                the rule.  Thus in the example above, we know that SPR,
                SPEC and SUBJ</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">need
                to be identified because the value of VAL is necessarily
                &quot;valence&quot;</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">(and
                not valence-min) as we&#39;ve mentioned COMPS. But if CASE
                were appropriate</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">for
                both noun and comp (for example), then we wouldn&#39;t be
                able to know to</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">put
                in identity<span> </span></span><span class="m_-8593124814449095464m_4777295698704568155m_-4850224290236657958m_-2966151228754498155gmail-il" style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">constraints</span><span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"><span> </span>for
                any other features of noun (or comp).  If</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">the
                daughter in fact had a<span> </span></span><span class="m_-8593124814449095464m_4777295698704568155m_-4850224290236657958m_-2966151228754498155gmail-il" style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">constraint</span><span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"><span> </span>on
                one of these other features,</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">it
                wouldn&#39;t </span><span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">be
                copied up to the mother.  Relatedly, we lose the actual
                HEAD </span></div>
            <div><span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">value
                because </span><span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">we
                can&#39;t identify HEAD while changing CASE.  (So here, the
                grammar </span></div>
            <div><span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">writer </span><span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">would
                need to stipulate [HEAD noun], say, on the mother.)</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">In
                Paris, I remember being convinced that the added
                simplicity in defining</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">lexical
                rules would out-weigh the lack of transparency noted
                above.  And</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">I&#39;m
                still pretty sure I agree with that.  One thing in favor
                of that </span><span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">view
                is that </span><span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">if
                a rule </span></div>
            <div><span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">defined
                using the<span> </span></span><span class="m_-8593124814449095464m_4777295698704568155m_-4850224290236657958m_-2966151228754498155gmail-il" style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">defeasible</span><span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"><span> </span>identity
                type didn&#39;t have the expected </span><span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">behavior,
                the </span></div>
            <div><span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">grammar
                engineer could always either add<span> </span></span><span class="m_-8593124814449095464m_4777295698704568155m_-4850224290236657958m_-2966151228754498155gmail-il" style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">constraints</span><span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"><span> </span>or
                side-step</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">that
                type and hand-specify all the desired identities.</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">A
                further complication I noticed while writing out this
                example is the</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">interaction
                between<span> </span></span><span class="m_-8593124814449095464m_4777295698704568155m_-4850224290236657958m_-2966151228754498155gmail-il" style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">defeasible</span><span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"><span> </span>and
                indefeasible identity tags.  Two conditions</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">to
                consider:</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">1)
                The rule inherits a<span> </span></span><span class="m_-8593124814449095464m_4777295698704568155m_-4850224290236657958m_-2966151228754498155gmail-il" style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">constraint</span><span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"><span> </span>(e.g.,
                from the type of the DTR value) that</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">the
                REST of the ARG-ST is the same as the COMPS list.</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">2)
                The rule doesn&#39;t inherit such a<span> </span></span><span class="m_-8593124814449095464m_4777295698704568155m_-4850224290236657958m_-2966151228754498155gmail-il" style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">constraint</span><span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">,
                but the constituent that</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">serves
                as the daughter identifies its ARG-ST.REST and its
                COMPS.</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">I
                think (2) isn&#39;t a problem (this is very similar to
                things that confused</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">Tom,
                Ivan and I as we designed the lex rules in the textbook,
                though, so I&#39;m</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">not
                feeling very confident just now!).  As for (1), it could
                entail a similar</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">push
                down of identity inside the ARG-ST.  But what if the
                ARG-ST to</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">DTR.ARG-ST
                identification were a non-</span><span class="m_-8593124814449095464m_4777295698704568155m_-4850224290236657958m_-2966151228754498155gmail-il" style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">defeasible</span><span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"><span> </span>identity<span> </span></span><span class="m_-8593124814449095464m_4777295698704568155m_-4850224290236657958m_-2966151228754498155gmail-il" style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">constraint</span><span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">?</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">Maybe
                that&#39;s just a broken grammar that either shouldn&#39;t
                compile or would</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">just
                have surprising behavior.</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">I
                hope you are still interested in this problem. Let me
                know if/when it would</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">be
                useful to have a grammar to play with.</span><br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <br style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial">
              <span style="font-size:12.8px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">Thanks!</span><br>
            </div>
          </div>
          <div class="gmail_extra"><br>
            <div class="gmail_quote">On Tue, Sep 4, 2018 at 10:42 AM, <a href="mailto:goodman.m.w@gmail.com" target="_blank">goodman.m.w@gmail.com</a> <span dir="ltr">&lt;<a href="mailto:goodman.m.w@gmail.com" target="_blank">goodman.m.w@gmail.com</a>&gt;</span>
              wrote:<br>
              <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
                <div dir="ltr">
                  <div dir="ltr">
                    <div dir="ltr">
                      <div dir="ltr">
                        <div style="font-family:arial,helvetica,sans-serif">Hello
                          everyone,</div>
                        <div style="font-family:arial,helvetica,sans-serif"><br>
                        </div>
                        <div style="font-family:arial,helvetica,sans-serif">I
                          appreciate the feedback I&#39;ve received in
                          previous messages in my attempts to dust off
                          neglected corners of TDL syntax, and I&#39;d now
                          like to bring up &quot;defaults&quot;, or &quot;defeasible
                          constraints&quot; (I believe these refer to the
                          same thing). Are we prepared to start
                          supporting defaults/defeasible-constraints in
                          our processors and using them in our grammars?
                          Or should we discard them as an undesired
                          experimental feature (i.e., declare them to
                          *not* be part of DELPH-IN TDL)?<br>
                          <div style="font-family:arial,helvetica,sans-serif"><br>
                          </div>
                          <div style="font-family:arial,helvetica,sans-serif">Further
                            information:<br>
                          </div>
                          <div style="font-family:arial,helvetica,sans-serif"><br>
                          </div>
                          Currently, only the LKB supports them (and
                          maybe PET?). As I understand, they are a
                          compile-time feature, meaning that they change
                          how the grammar is compiled and that there is
                          no longer a notion of &quot;defaults&quot; during
                          run-time. I don&#39;t think the use of defaults
                          causes any change in the competence or
                          performance of a grammar.<br>
                        </div>
                        <div style="font-family:arial,helvetica,sans-serif"><br>
                        </div>
                        <div style="font-family:arial,helvetica,sans-serif">The
                          benefit of defaults is for the grammar
                          engineer as it can reduce the amount of
                          boilerplate code and make the grammar source
                          code more intuitive. I think any result that
                          makes grammar writing easier is a big win. The
                          differences it creates between the source-code
                          form of the grammar and the compiled
                          hierarchy, however, can complicate debugging
                          (e.g., interactive unification).<br>
                        </div>
                        <div style="font-family:arial,helvetica,sans-serif"><br>
                        </div>
                        <div style="font-family:arial,helvetica,sans-serif">Some
                          links:</div>
                        <div style="font-family:arial,helvetica,sans-serif"> 
                          - <a href="http://www.aclweb.org/anthology/J99-1002" target="_blank">http://www.aclweb.org/anthology/J99-1002</a></div>
                        <div style="font-family:arial,helvetica,sans-serif"> 
                          - <a href="http://moin.delph-in.net/ParisDefeasibleConstraints" target="_blank">http://moin.delph-in.net/ParisDefeasibleConstraints</a></div>
                        <div style="font-family:arial,helvetica,sans-serif"> 
                          - <a href="http://moin.delph-in.net/StanfordDefaults" target="_blank">http://moin.delph-in.net/StanfordDefaults</a><span class="m_-8593124814449095464m_4777295698704568155m_-4850224290236657958m_-2966151228754498155HOEnZb"><font color="#888888"><br clear="all">
                            </font></span></div>
                        <span class="m_-8593124814449095464m_4777295698704568155m_-4850224290236657958m_-2966151228754498155HOEnZb"><font color="#888888"><br>
                            -- <br>
                            <div dir="ltr" class="m_-8593124814449095464m_4777295698704568155m_-4850224290236657958m_-2966151228754498155m_-2873427553648370275gmail_signature">-Michael
                              Wayne Goodman</div>
                          </font></span></div>
                    </div>
                  </div>
                </div>
              </blockquote>
            </div>
            <br>
            <br clear="all">
            <div><br>
            </div>
            -- <br>
            <div class="m_-8593124814449095464m_4777295698704568155m_-4850224290236657958m_-2966151228754498155gmail_signature" data-smartmail="gmail_signature">
              <div dir="ltr">
                <div>
                  <div dir="ltr">
                    <div>
                      <div dir="ltr">
                        <div>
                          <div dir="ltr">
                            <div>
                              <div dir="ltr">Emily M. Bender<br>
                                Professor, <span style="font-size:12.8px">Department of
                                  Linguistics</span></div>
                              <div><span style="font-size:12.8px">University
                                  of Washington</span></div>
                              <div>Twitter: @emilymbender</div>
                            </div>
                          </div>
                        </div>
                      </div>
                    </div>
                  </div>
                </div>
              </div>
            </div>
          </div>
        </blockquote>
      </div>
    </blockquote>
    <br>
  </div>

</blockquote></div></div></div>
</blockquote></div><br clear="all"><br>-- <br><div dir="ltr" class="m_-8593124814449095464gmail_signature" data-smartmail="gmail_signature">-Michael Wayne Goodman</div>
</blockquote></div></div>