this is for holding javascript data
test1234567 deleted Id_nulls_tex_v_1__.tex
about 9 years ago
Commit id: 6b5f537f3a7ba7bb80572fe2a9ce2b001da4e7cf
deletions | additions
diff --git a/Id_nulls_tex_v_1__.tex b/Id_nulls_tex_v_1__.tex
deleted file mode 100644
index eceb3c2..0000000
--- a/Id_nulls_tex_v_1__.tex
+++ /dev/null
...
% $Id: nulls.tex,v 1.40 1994/12/20 07:24:21 robertm Exp $
%\input{stuff}
%\makeindex
%\begin{document}
\section{Null Elements}
\label{list:nulls}
%I.
\subsection{The building blocks} \index{null elements|(}
%A.
\subsubsection{Inventory}
The inventory of null elements is
\begin{codelist}
\item[*T*] (trace of A$^\prime$-movement, including parasitic gaps)
\item[(NP *)] (arbitrary PRO, controlled PRO, and trace of A-movement)
\item[0] (null complementizer, including null \iwh-operator)
\item[*U*] (unit)
\item[*?*] (placeholder for ellipsed material)
\item[*NOT*] (anti-placeholder in template gapping)
\item[*RNR*] (pseudo-attach: right node raising)
\item[*ICH*] (pseudo-attach: interpret constituent here)
\item[*EXP*] (pseudo-attach: expletive)
\item[*PPA*] (pseudo-attach: permanent predictable ambiguity)
\end{codelist}
But see also section~\ref{list:nulls:illegal} for list of illegal null
elements that may also appear.
Note that while most null elements contain \code{*}'s, they are not the
only asterisks in the texts; there are also a few naturally-occurring
footnote markers. In the ``combined'' files, null elements are tagged
-NONE-, while footnote markers are generally tagged SYM and preceded by a
backslash.
%B.
\subsubsection{Indexing} \index{coindexing|(} \index{indexing|(}
Indices are used only when they can be used to indicate a relationship
that would otherwise not be unambiguously retrievable from the bracketing.
Indices are used to express such relationships as coreference (as in the
case of controlled PRO or pragmatic coreference for arbitrary PRO), binding
(as in the case of \iwh-movement), or close association (as in the
case of {\it it}\/-extraposition). These relationships are shown only when
some type of null element is involved, and only when the relationship is
intrasentential. One null element may be associated with another, as in
the case of the null \iwh-operator. Coreference relations between overt
pronouns and their antecedents are not annotated.
\paragraph{The identity index.} \index{identity index}
In principle, each bracket within the topmost S is understood to have a
unique index (an ``identity index''), which in practice is used only when
that constituent is coreferent with or otherwise closely associated with
some null element in the sentence (or when it's acting in a gapping
``template''). The brackets surrounding null elements are also understood
to be associated with a unique identity index. Identity indices appear
only on the bracket label, as in (NP-1 Kris), (WHNP-2 which dog), (SBAR-24
who offered to take me for a swim), etc.
The actual numbering of the identity indices is arbitrary; i.e. the
constituents are not necessarily numbered sequentially within the sentence,
and a given sentence may contain brackets with the identity indices -1, -2,
-5, and -1978. Note also that in rare cases, a bracket may have an
identity index shown when there is no corresponding null element or
template-gapping constituent.
\paragraph{The reference index.} \index{reference index}
In most cases, a null element will be suffixed
with an integer (the ``reference index'') that matches the identity index
on the bracket label of some other constituent. Note that the reference
index on the null element takes the form of a dash-number on the null
element itself, and not on the bracket label, as in (NP *-1), (NP *T*-2),
and (SBAR *ICH*-24). If the null element in turn refers to or is
associated with a third element, it will bear its own identity index, along
the lines of (NP-1 *T*-2).
\begin{verbatim}
(S (NP-SBJ-1 he)
(VP was
(VP accused
(NP-3 *-1)
(PP-CLR of
(S-NOM (NP-SBJ *-3)
(VP (VP conducting
(NP illegal business))
and
(VP possessing
(NP illegal materials))))))))
\end{verbatim}
\begin{verbatim}
(S (NP-SBJ (NP It)
(S *EXP*-2))
(VP 's
(ADJP-PRD (ADJP easier)
(SBAR *ICH*-1))
(S-2 (NP-SBJ *)
(VP to
(VP get
(ADJP-PRD worse))))
(SBAR-1 than
(FRAG (ADJP-PRD better)))
(PP-LOC in
(NP this game))))
\end{verbatim}
% I think this was never said definitively one way or the other, but are we
% supposed to use -PRD inside a FRAG? I thought we weren't supposed to be
% so smart inside FRAG, even though we know we are. -mark.
(Note that while this indexing system appears complex, the annotation
procedure is actually quite simple, as it is accomplished with simple mouse
drags that make attention to the gritty details of coindexation unnecessary.)
Indices are also used for the various kinds of pseudo-attach (described in
\RefPseudo) and for template gapping (described in \RefCoord).
\index{coindexing|)} \index{indexing|)}
%C.
\subsubsection{Other tags}
Null elements may bear additional function tags, as described in
\RefNotation. For instance, the grammatical function of extracted {\it
wh}\/-phrases is noted on the \iwh-trace and not on the {\it
wh}\/-element itself.
\begin{verbatim}
(SBARQ (WHADVP-439 Where)
(SQ did
(NP-SBJ you)
(VP put
(NP the book)
(ADVP-PUT *T*-439)))
?)
\end{verbatim}
%II.
\subsection{*T* (trace of A$^\prime$ movement)} \index{T@*T*|(}
Although the use of *T* corresponds loosely to A$^\prime$ movement, the
match is not precise (e.g. it includes ``parasitic gaps''). *T* can also
be seen as marking the interpretation location of certain constituents that
are not in their usual argument position.
The trace *T* always bears a referential index that corresponds to the
identity index of some other constituent in the sentence (moved {\it
wh}\/-word, topicalized NP or ADVP, etc.).
%A.
\subsubsection{{\it Wh}\/-questions} \index{T@*T*!{\it wh}\/-questions}
\index{wh-questions@{\it wh}\/-questions!*T*}
% Some of the examples here may also be covered in the WH-phrases section.
% If we have time, someone could go through and try to eliminate the
% redundancies. In the meantime, this'll do.
{\it Wh}\/-moved noun phrases are labeled WHNP and put inside SBARQ. They
bear an identity index that matches the reference index on the *T* in the
position of the gap. Constituents other than NP are labeled WHxx (WHADVP,
WHPP, or WHADJP, as appropriate), placed under SBARQ, and coindexed with
the *T* in the position of the gap. The same procedure holds for both
arguments and adjuncts.
\begin{itemize}
\item NP arguments
\begin{verbatim}
(SBARQ (WHNP-1 what)
(SQ are
(NP-SBJ you)
(VP thinking
(PP-CLR about
(NP *T*-1))))
?)
\end{verbatim}
% [a note on parentheses for posterity]
% Note that outermost unlabelled ()'s, while not terribly important for the
% public manual, are probably necessary for the Training Manual, since
% otherwise it's confusing to the annotators when they see real data files.
% (Notably, final punctuation has two close parentheses after it in
% properly coded data files.) One annotator in the deep dark past (hudson)
% ended up destroying the integrity of about 100 files because of such
% confusion.
%
% Also, care should be taken when removing ()'s not to create unbalanced
% parentheses, since they can be both misleading to the reader and
% potentially dangerous when automatic formatting tools are used.
\begin{verbatim}
(SBARQ (WHNP-1 (WHNP Which story)
(PP about
(NP tribbles)))
(SQ did
(NP-SBJ you)
(VP read
(NP *T*-1)))
?)
\end{verbatim}
\begin{verbatim}
(SBARQ (WHNP-1 what time)
(SQ is
(NP-SBJ it)
(NP-PRD *T*-1))
?)
\end{verbatim}
\item NP adjunct
\begin{verbatim}
(SBARQ (WHNP-1 Which day)
(SQ did
(NP-SBJ you)
(VP get
(ADVP-DIR there)
(NP-TMP *T*-1)))
?)
\end{verbatim}
\item Non-NP arguments
\begin{verbatim}
(SBARQ (WHADVP-439 Where)
(SQ did
(NP-SBJ you)
(VP put
(NP the book)
(ADVP-PUT *T*-439)))
?)
\end{verbatim}
\begin{verbatim}
(SBARQ (WHPP-42 On
(WHNP what))
(SQ did
(NP-SBJ you)
(VP sit
(PP-LOC-CLR *T*-42)))
?)
\end{verbatim}
\begin{verbatim}
(SBARQ (WHADJP-54 How cold)
(SQ is
(NP-SBJ it)
(ADJP-PRD *T*-54)
(ADVP-LOC outside))
?)
\end{verbatim}
\item Non-NP adjuncts
\begin{verbatim}
(SBARQ (WHADVP-42 How)
(SQ did
(NP-SBJ you)
(VP fix
(NP the car)
(ADVP-MNR *T*-42)))
?)
\end{verbatim}
\begin{verbatim}
(SBARQ (WHADVP-1 Where)
(SQ did
(NP-SBJ you)
(VP meet
(NP them)
(ADVP-LOC *T*-1)))
?)
\end{verbatim}
\begin{verbatim}
(SBARQ (WHADVP-54 Why)
(SQ did
(NP-SBJ you)
(VP jump
(PP-DIR off
(NP the cliff))
(ADVP-PRP *T*-54)))
?)
\end{verbatim}
\end{itemize}
%B.
\subsubsection{Relative clauses} \index{T@*T*!relative clauses}
\index{relative clauses!T@*T*}
Relative clauses are adjoined to the head noun phrase. The relative
pronoun is given the appropriate WH-label, put inside the SBAR level, and
coindexed with a *T* in the position of the gap. (Note that relative
clauses differ from (direct) \iwh-questions in that they contain an SBAR
rather than an SBARQ.)
% 1.
\paragraph{\iwh- and ``that'' relative clauses.}
\index{wh-clauses@{\it wh}\/-clauses!relative clauses}
Relative clauses introduced by {\it that}\/ are annotated just as relative
clauses introduced by a \iwh-word: {\it that}\/ is given the
appropriate WH-label, put inside an SBAR level, and coindexed with the *T*
in the position of the gap.
\begin{itemize}
\item NP trace
\begin{verbatim}
(NP (NP answers)
(SBAR (WHNP-6 that/which)
(S (NP-SBJ-3 we)
(VP 'd
(VP like
(S (NP-SBJ *-3)
(VP to
(VP have
(NP *T*-6)))))))))
\end{verbatim}
\item ADVP trace
\begin{verbatim}
(NP (NP the place)
(SBAR (WHADVP-2 that/where)
(S (NP-SBJ I)
(VP put
(NP the book)
(ADVP-PUT *T*-2)))))
\end{verbatim}
\end{itemize}
%2.
\paragraph{Zero relatives.} \index{relative clauses!zero relatives}
Relative clauses introduced by a null complementizer are annotated in a
similar fashion, this time with a null complementizer `0' inside SBAR
labeled with the appropriate \iwh-category and coindexed with a *T*
in the position of the gap.
\begin{itemize}
\item NP trace
\begin{verbatim}
(NP (NP answers)
(SBAR (WHNP-3 0)
(S (NP-SBJ-4 we)
(VP 'd
(VP like
(S (NP-SBJ *-4)
(VP to
(VP have
(NP *T*-3)))))))))
\end{verbatim}
\item ADVP trace
\begin{verbatim}
(NP (NP the place)
(SBAR (WHADVP-2 0)
(S (NP-SBJ I)
(VP put
(NP the book)
(ADVP-PUT *T*-2)))))
\end{verbatim}
\end{itemize}
%3.
\paragraph{Infinitival relatives.} See \RefInf\ for more information.
\index{infinitival relatives}
\begin{itemize}
\item trace as object
\begin{verbatim}
(NP (NP a movie)
(SBAR (WHNP-1 0)
(S (NP-SBJ *)
(VP to
(VP see
(NP *T*-1))))))
\end{verbatim}
\item trace as subject
\begin{verbatim}
(NP (NP bloodhounds)
(SBAR (WHNP-4 0)
(S (NP-SBJ *T*-4)
(VP to
(VP trail
(NP the assassins))))))
\end{verbatim}
\item trace as adjunct
\begin{verbatim}
(NP (NP time)
(SBAR (WHADVP-1 0)
(S (NP-SBJ *)
(VP to
(VP go
(ADVP-TMP *T*-1))))))
\end{verbatim}
\end{itemize}
%C.
\subsubsection{Fronted elements} \index{T@*T*!fronted elements}
Fronted elements are placed inside the top clause level (e.g. S, SINV,
SQ, SBAR). (Only certain fronted elements are tagged -TPC: (i)
constituents associated with a *T* in the position of the gap and (ii)
left-dislocated constituents (those associated with a resumptive pronoun in
the position of the gap).) (See \RefOverview\ for more details on the
treatment of fronted elements.) \index{TPC@-TPC}
% 1.
\paragraph {Arguments.}~
Fronted argument noun phrases are coindexed with a *T* in the position of
the gap:
\begin{verbatim}
(S (NP-TPC-3 This)
(NP-SBJ every man)
(VP contains
(NP *T*-3)
(PP-LOC-CLR within
(NP him))))
\end{verbatim}
\begin{verbatim}
(S (NP-TPC-4 Our dull unsystematic youth)
(NP-SBJ we)
(VP let
(S (NP-SBJ *T*-4)
(VP stray
(PP-DIR into
(NP philanthropy))))))
\end{verbatim}
If the fronted argument is an instance of left-dislocation (i.e, associated
with a resumptive pronoun), there is no coindexation between the fronted
argument and the pronoun: \index{left-dislocation}
\begin{verbatim}
(S (NP-TPC John)
,
(NP-SBJ I)
(VP like
(NP him)
(NP-ADV a lot)))
\end{verbatim}
Other fronted arguments (such as the main VP, a predicate, the locative
complement of {\it put}\/, etc.) are also tagged -TPC, and their identity
index matches the reference index on the *T* inserted in the position of
the gap. \index{TPC@-TPC}
\begin{verbatim}
(S (ADVP-PUT-TPC-1 There)
,
(NP-SBJ I)
(VP put
(NP the book)
(ADVP-PUT *T*-1)))
\end{verbatim}
\begin{verbatim}
(S (SBAR-ADV (VP-TPC-2 Shout
(PP-CLR at
(NP Eichmann)))
though
(S (NP-SBJ he)
(VP might
(VP *T*-2))))
the prosecutor could not establish...)
\end{verbatim}
\begin{verbatim}
(S (SBAR-ADV (ADJP-PRD-TPC-5 Wrong)
though
(S (NP-SBJ the policy)
(VP may
(VP be
(ADJP-PRD *T*-5)))))
it at least works pretty often.)
\end{verbatim}
\begin{verbatim}
(S (S (NP-SBJ-1 we)
(VP hope
(S (NP-SBJ *-1)
(VP to
(VP have
(NP a million dollars)
(NP-TMP someday))))))
and
(S (VP-TPC-6 have
(NP it)
(SBAR-ADV *ICH*-2))
(NP-SBJ we)
(VP may
(VP *T*-6)
,
(SBAR-ADV-2 if
(S (NP-SBJ we)
(VP get
(ADJP-PRD lucky)))))))
\end{verbatim}
Quotations that precede a verb of saying are treated as fronted arguments:
they leave a *T* and receive the -TPC tag. (See \RefOverview\ for more
details on the treatment of quotations.) \index{quotations} \index{TPC@-TPC}
\begin{verbatim}
( (S ``
(S-TPC-1 (NP-SBJ We)
(VP will
(VP win)))
,
''
(NP-SBJ Mary)
(VP said
(S *T*-1))
.))
\end{verbatim}
Note that any constituent tagged -CLR is considered an argument for these
purposes: it leaves a *T* and receives the -TPC tag if fronted.
\begin{verbatim}
(S (PP-CLR-TPC-5 With
(NP final exams))
,
(NP-SBJ I)
(VP associate
(NP blood , sweat , and tears)
(PP-CLR *T*-5)))
\end{verbatim}
% 2.
\paragraph{Adjuncts.}~
Fronted adjuncts are not associated with a *T* when they have not left the
clause in which they originate, since in this case their relation to the
clause is still clear.
\begin{verbatim}
(S (ADVP-MNR Carefully)
,
(NP-SBJ I)
(VP dropped
(NP the feathers)))
\end{verbatim}
\begin{verbatim}
(S (NP-TMP Yesterday)
(NP-SBJ-1 a child)
(VP came
(ADVP-DIR out)
(S-PRP (NP-SBJ *-1)
(VP to
(VP wonder)))))
\end{verbatim}
\begin{verbatim}
(SBARQ (NP-TMP Yesterday)
,
(WHNP-2 what)
(SQ did
(NP-SBJ we)
(VP decide
(NP *T*-2)
(PP for
(NP this one))))
?)
\end{verbatim}
However, adjuncts that originate in a lower clause are associated with a
*T* in the position of the gap.
\begin{verbatim}
(S (SBAR-PRP-TPC-9 Because
(S (NP-SBJ I)
(VP 'm
(NP-PRD such a bad boy))))
(NP-SBJ I)
(VP think
(SBAR 0
(S (NP-SBJ I)
(VP wo n't
(VP get
(NP a lollipop)
(SBAR-PRP *T*-9)))))))
\end{verbatim}
In cases where it is ambiguous whether the adjunct originates in a lower
clause or in the matrix clause, the adjunct is analyzed as originating from
the matrix clause and NOT bracketed -TPC.
%D.
\subsubsection{Tough movement} \index{tough movement}
The null element is coindexed to a null \iwh-phrase.
See \RefInf\ for more information.
\begin{verbatim}
(S (NP-SBJ Cars)
(VP are
(ADJP-PRD tough
(SBAR (WHNP-7 0)
(S (NP-SBJ *)
(VP to
(VP pay
(PP-CLR for
(NP *T*-7)))))))))
\end{verbatim}
%E.
\subsubsection{Parasitic gaps} \index{parasitic gap}
A coindexed *T* is put in the parasitic gap as well as in the position of
the original gap.
\begin{verbatim}
(SBARQ (WHNP-1 Which papers)
(SQ did
(NP-SBJ-2 you)
(VP file
(NP *T*-1)
(PP without
(S-NOM (NP-SBJ *-2)
(VP reading (NP *T*-1))))))
?)
\end{verbatim}
\index{T@*T*|)}
%III.
\subsection{\code{*} (trace of NP movement, controlled PRO, arbitrary PRO)}
Because it corresponds to the trace of NP movement, controlled PRO, or
arbitrary PRO, \code{*} always appears within NP: \code{(NP *)}.
\index{PRO!controlled} \index{PRO!arbitrary}
%A.
\subsubsection{Indexing}
(NP *) bears a reference index whenever it is fairly clear what nominal it
is controlled by, corresponding roughly to controlled PRO and the passive
trace. However, indexing also reflects pragmatic coreference in addition
to syntactic relations, within limits described below and in
section~\ref{list:nulls:coindex-limits}.
Unlike *T*, * may appear without a reference index. Unindexed (NP *)
corresponds roughly to arbitrary PRO (and passive traces appearing in
Reduced Relative Clauses---see section~\ref{list:nulls:RRCs}).
In cases of strings of coindexed null elements, the null is coindexed to
the most local NP, as with with passives under raising predicates or in the
following example:
\begin{verbatim}
(S (NP-SBJ-1 he)
(VP was
(VP accused
(NP-3 *-1)
(PP-CLR of
(S-NOM (NP-SBJ *-3)
(VP (VP conducting
(NP illegal business))
and
(VP possessing
(NP illegal materials))))))))
\end{verbatim}
When several NPs are adjoined, the indexing should be from the highest
NP:
% from wsj_0019
\begin{verbatim}
(S (NP-SBJ-28 (NP Arthur A. Hatch)
,
(NP 59)
,)
(VP was
(VP named
(S (NP-SBJ *-28)
(NP-PRD (NP executive vice president)
(PP of
(NP the company)))))))
\end{verbatim}
%B.
\subsubsection{Passives}
% 1.
\paragraph{Object of verb.}
The trace (NP *) is put after the passive verb and
coindexed with the constituent in subject position.
\begin{verbatim}
(S (NP-SBJ-1 John)
(VP was
(VP hit
(NP *-1)
(PP by
(NP-LGS a ball)))))
\end{verbatim}
Note that the * may come before or after a PRT (particle). There is no
policy governing this and either order is possible, though it is somewhat
more likely for the PRT to come second: \index{PRT}
% tgrep sez: 178 PRT second, 68 PRT first
\begin{verbatim}
(S (NP-SBJ-1 Arthur)
(VP was
(VP picked
(NP *-1)
(PRT up)
(PP by
(NP-LGS aliens)))))
\end{verbatim}
\begin{verbatim}
(S (NP-SBJ-1 Arthur)
(VP was
(VP picked
(PRT up)
(NP *-1)
(PP by
(NP-LGS aliens)))))
\end{verbatim}
% 2.
\paragraph{Object of preposition.}
The null may be put after prepositions as
required by the construction.
\begin{verbatim}
(S (NP-SBJ-2 (NP kid 's) cars)
(VP are
(ADVP-TMP often)
(VP paid
(PP-CLR for
(NP *-2))
(PP by
(NP-LGS their parents)))))
\end{verbatim}
% 4.
\paragraph{In reduced relative clause.}
See section~\ref{list:nulls:RRCs} for more details on the
treatment of reduced relative clauses. The passive verb (or preposition,
as the case may be) in a reduced relative clause is also followed by a
(NP *). The passive trace in a reduced relative is not coindexed to the
NP preceding it. This reflects an understanding of the relationship
between the NP and reduced relative as post-modification rather than
predication.
\begin{verbatim}
(NP (NP an agreement)
(VP signed
(NP *)
(PP by
(NP-LGS everyone))))
\end{verbatim}
\begin{verbatim}
(NP (NP a government service)
(VP paid
(PP-CLR for
(NP *))
(PP by
(NP-LGS everyone))))
\end{verbatim}
In some cases (NP *) may function as the subject of a small clause within
the reduced relative:
\begin{verbatim}
(NP (NP an elephant)
(VP called
(S (NP-SBJ *)
(NP-PRD Dumbo))))
\end{verbatim}
%C.
\subsubsection{Subjects of participial clauses and gerunds}
The null subject (NP *) of a participial clause or gerund is coindexed
with another constituent in the sentence if it is clear to the annotator
that the two are coreferent. No attempt is made to conform to the standard
Binding Theory of GB or any other such formal approach to coreference. See
\RefGP\ for more details on the annotation of participial clauses and
gerunds.
\begin{itemize}
\item VP complements
\begin{verbatim}
(S (NP-SBJ-1 I)
(VP stopped
(S (NP-SBJ *-1)
(VP eating
(NP chocolate)))
(PP-PRP for
(NP Lent))))
\end{verbatim}
\begin{verbatim}
(S But
(NP-SBJ-1 I)
(VP liked
(S (NP-SBJ *-1)
(VP eating
(NP chocolate)))
(ADVP-TMP before)))
\end{verbatim}
\begin{verbatim}
(S (NP-SBJ-1 he)
(VP was
(VP accused
(NP-3 *-1)
(PP-CLR of
(S-NOM (NP-SBJ *-3)
(VP (VP conducting
(NP illegal business))
and
(VP possessing
(NP illegal materials))))))))
\end{verbatim}
\item Adverbials
\begin{verbatim}
(S (NP-SBJ-1 She)
(VP left
,
(S-ADV (NP-SBJ-2 *-1)
(VP offended
(NP *-2)
(PP by (NP-LGS their remarks))))))
\end{verbatim}
%this one's pretty funny:
\begin{verbatim}
(S (NP-SBJ-1 Time)
(VP eluded
(NP Paramount)
(PP-MNR by
(S-NOM (NP-SBJ *-1)
(VP acquiring
(NP Warner Communications Inc))))))
\end{verbatim}
\begin{verbatim}
(S (NP-SBJ-1 I)
(VP fell
(S-CLR (NP-SBJ *-1)
(ADJP-PRD asleep))
(PP-LOC on
(NP the lobby floor))))
\end{verbatim}
\begin{verbatim}
(S (NP-SBJ-1 Borough Presidents)
,
(SBAR-ADV while
(S (NP-SBJ *-1)
(VP retaining
(NP (NP membership)
(PP-LOC in
(NP (NP the Board)
(PP of
(NP Estimate))))))))
,
(VP lose
(NP their housekeeping functions)))
\end{verbatim}
\item Subjects
\begin{verbatim}
(S (S-NOM-SBJ (NP-SBJ *)
(VP Eating
(NP chocolate)))
(VP is
(ADJP-PRD good
(PP for
(NP you)))))
\end{verbatim}
\item Without coindexation.
In the case where there is no good candidate for coreference within the
sentence, (NP *) remains without an index.
\begin{verbatim}
(S (NP-SBJ A Texas legislator)
(VP proposes
(S (NP-SBJ *)
(VP color-coding
(NP (NP (NP drivers ')
licenses)
(PP of
(NP some drug offenders)))))))
\end{verbatim}
\begin{verbatim}
(S (S-NOM-SBJ (NP-SBJ *)
(VP Taking
(NP Iwo Jima)))
(VP was
(NP-PRD no easy feat)))
\end{verbatim}
Note also that the null subject of a gerund that is coordinated with one or
more NPs is usually not coindexed.
\begin{verbatim}
(S (NP-SBJ I)
(VP like
(NP (NP medals)
,
(NP cheering crowds)
,
and
(S-NOM (NP-SBJ *)
(VP swimming
(NP the backstroke))))))
\end{verbatim}
\end{itemize}
%D.
\subsubsection{Subjects of infinitival clauses}
% 1.
\paragraph{With coindexation}~
\begin{enumerate}
% a.
\item VP complement clauses.
Note that from the perspective of the
annotator, it is not necessary to distinguish between Raising and Control
structures, etc. In each case, the annotator simply coindexes the empty
subject of the infinitival with whatever lexical NP it is associated with.
\begin{enumerate}
% i.
\item ``Raising'' constructions.
\begin{verbatim}
(S (NP-SBJ-3 Everyone)
(VP seems
(S (NP-SBJ *-3)
(VP to
(VP dislike
(NP Drew Barrymore))))))
\end{verbatim}
% ii.
\item ``Object control'' constructions.
\begin{verbatim}
(S (NP-SBJ Ford)
(VP persuaded
(NP-1 Zaphod)
(S (NP-SBJ *-1)
(VP to
(VP run
(PP-CLR for
(NP president)))))))
\end{verbatim}
% iii.
\item ``Subject control'' constructions.
\begin{verbatim}
(S (NP-SBJ-1 Zaphod)
(VP promised
(NP Ford)
(S (NP-SBJ *-1)
(VP to
(VP run
(PP-CLR for
(NP president)))))))
\end{verbatim}
% iv.
\item Semi-auxiliaries. \index{semi-auxiliaries}
Semi-auxiliaries occur in constructions with infinitival {\it to}\/,
(e.g, {\it supposed to, ought to, have to}\/). They are annotated with
full infinitival structure and have a (NP-SBJ *) subject, coindexed as
appropriate.
\begin{verbatim}
(S (PP Of (NP course))
,
(NP-SBJ-1 regulators)
(VP would
(VP have
(S (NP-SBJ *-1)
(VP to
(VP approve
(NP (NP Columbia 's)
reorganization)))))))
\end{verbatim}
{\it about to} \index{about to@{\it about to}} is also treated as a
semi-auxiliary in official policy, though some variation exists.
The following bracketings are likely.
\begin{itemize}
\item Official Policy:
% I replaced the original with a simpler, real example.-R.
\begin{verbatim}
(S (NP-SBJ-2 interest rates)
(VP are
(VP about
(S (NP-SBJ *-2)
(VP to
(VP nosedive))))))
\end{verbatim}
\item Common alternate:
\begin{verbatim}
(S (NP-SBJ-1 the pollen)
(VP is
(ADJP-PRD about
(S (NP-SBJ *-1)
(VP to
(VP develop))))))
\end{verbatim}
\end{itemize}
% Adjunct to NP:
%
% ( (SQ Does
% (NP-SBJ the tyke)
% (VP have
% (NP (NP a good mind)
% (ADJP about
% (S (NP-SBJ-1 *)
% (VP to
% (VP be
% (VP wasted
% (NP *-1))))))))
% ?))
\end{enumerate}
% b.
\item ADJP complement clauses.
Null element is coindexed with the matrix subject, where appropriate.
\begin{enumerate}
% i.
\item ``Raising'' constructions.
\begin{verbatim}
(S (NP-SBJ-4 This climb)
(VP is
(ADJP-PRD likely
(S (NP-SBJ *-4)
(VP to
(VP be
(ADJP-PRD difficult)))))))
\end{verbatim}
% ii.
\item ``Control'' constructions.
\begin{verbatim}
(S (NP-SBJ-1 Zaphod)
(VP is
(ADJP-PRD ready
(S (NP-SBJ *-1)
(VP to
(VP eat
(NP the steak)))))))
\end{verbatim}
\end{enumerate}
% c.
\item Adverbials.
Here are a few common adverbials infinitives. This is not an exhaustive list.
\begin{enumerate}
\item purpose clause
\begin{verbatim}
(S (NP-SBJ-1 The public)
(VP did n't
(VP come
(PP-DIR to
(NP the market))
(S-PRP (NP-SBJ *-1)
(VP to
(VP play
(NP a game)))))))
\end{verbatim}
\item semi-complement clause
\begin{verbatim}
(S (NP-SBJ-1 Skilled ringers)
(VP use
(NP their wrists)
(S-CLR (NP-SBJ *-1)
(VP to
(VP advance or retard
(NP the next swing))))))
\end{verbatim}
\item resultative clause
\begin{verbatim}
(S (NP-SBJ-1 (NP London 's)
Financial Times 100-share index)
(VP shed
(NP 40.4 points)
(S-ADV (NP-SBJ *-1)
(VP to
(VP finish
(PP-CLR at
(NP 2149.3)))))))
\end{verbatim}
\end{enumerate}
\end{enumerate}
% 2.
\paragraph{Without coindexation.}
The following types of (NP *) subject are not
coindexed: subject of infinitive inside NP, imperative subject, subject in
``tough-movement'' construction. Aside from these rules, there are some
cases in which coindexation is much less likely than normal, described in
section~\ref{list:nulls:coindex-limits}.
\begin{enumerate}
% a.
\item Infinitives inside NPs.
These include complement clauses within NP
and the subjects of infinitival relative clauses,
\begin{verbatim}
(NP (NP John 's)
decision
(S (NP-SBJ *)
(VP to
(VP leave))))
\end{verbatim}
\begin{verbatim}
(S (NP-SBJ I)
(VP made
(NP a decision
(S (NP-SBJ *)
(VP to
(VP leave))))))
\end{verbatim}
\begin{verbatim}
(NP (NP a manual)
(SBAR (WHNP-1 0)
(S (NP-SBJ *)
(VP to
(VP write
(NP *T*-1))))))
\end{verbatim}
Note, however, that the *T* subject of an infinitival relative is coindexed
as appropriate.
\begin{verbatim}
(NP (NP bloodhounds)
(SBAR (WHNP-4 0)
(S (NP-SBJ *T*-4)
(VP to
(VP trail
(NP the assassins))))))
\end{verbatim}
% b.
\item Imperative subjects.
\begin{verbatim}
(S (NP-VOC Kris)
,
(NP-SBJ *)
(VP go
(ADVP-DIR home))
!)
\end{verbatim}
% c.
\item Subjects in ``tough''-movement constructions. \index{tough
movement}
\begin{verbatim}
(S (PP for (NP Zaphod))
,
(NP-SBJ that steak)
(VP is
(ADJP-PRD ready
(SBAR (WHNP-1 0)
(S (NP-SBJ *)
(VP to
(VP eat
(NP *T*-1))))))))
\end{verbatim}
\end{enumerate}
%E.
\subsubsection{Subjects of {\it as}\/- and {\it than}\/-clauses}
A (NP *) is used as a ``placeholder'' subject in clauses introduced by {\it
than}\/ or {\it as}\/ that lack an overt subject. This type of structure
may also be annotated using *?* or FRAG, though both are rare. See then
end of section~\ref{list:nulls:AsThan-alternatives}
(page~\pageref{list:nulls:AsThan-alternatives}) for more on these
possibilities.
\begin{verbatim}
(S But
(NP-SBJ there)
(VP may
(VP be
(NP-PRD (NP less)
(ADVP-LOC there)
(SBAR than
(S (NP-SBJ *)
(VP meets
(NP the eye))))))))
\end{verbatim}
% shouldn't the (ADVP-LOC there) go a level higher? -m.
% I think attaching it higher would result in recursive NP adjunction. -k.
% You can think ``VP adjunct'' -> ICH-attach the ``than''. Or, as is done
% here, you can think ``NP adjunct''. Depends on your interpretation.-R.
\begin{verbatim}
(NP (NP as little)
(SBAR as
(S (NP-SBJ *)
(VP is
(ADJP-PRD consistent ...)))))
\end{verbatim}
\begin{verbatim}
(S (NP-SBJ Primerica)
,
(SBAR-ADV as
(S (NP-SBJ *)
(VP expected)))
,
(ADVP also)
(VP acquired ...))
\end{verbatim}
% added passive trace to ``as expected'' clause... guess it could also be,
% alternately, a (S *?*), as I think is pointed out above. ?-m.
% It's unclear to me that anything should go there, especially a passive
% trace. In any case, the actual example (in wsj_0150) has neither * nor
% *?*.-R.
\begin{verbatim}
(NP (NP Items)
(VP listed
(NP-1 *)
(PP-CLR as
(S-NOM (NP-SBJ *-1)
(VP being
(PP-PRD in
(NP short supply)))))))
\end{verbatim}
\begin{verbatim}
(S (NP-SBJ-1 She)
(VP was
(VP quoted
(NP-2 *-1)
(PP-CLR as
(S-NOM (NP-SBJ *-2)
(VP saying ...))))))
\end{verbatim}
%IV.
\subsection{0 (null complementizer)} \index{zero@0 (null complementizer)|(}
\code{0} is used inside SBAR only if there is no overt \iwh-element
or {\it that}\/: \code{(SBAR 0)}\/ or \code{(SBAR (WHxx 0))}\/.
%A.
\subsubsection{Subordinator for tensed complement clauses}
The null complementizer introduces most tensed complement clauses.
% 1.
\begin{itemize}
\item With complement of ADJP.
\begin{verbatim}
(S (NP-SBJ I)
(VP 'm
(ADJP-PRD sure
(SBAR 0
(S (NP-SBJ he)
(VP 'll
(VP be
(ADVP-LOC-PRD here)
(NP-TMP any minute))))))))
\end{verbatim}
% 2.
\item With complement of VP.
\begin{verbatim}
(S (NP-SBJ I)
(VP believe
(SBAR 0
(S (NP-SBJ you)
(VP are
(ADJP-PRD *?*))))))
\end{verbatim}
% 3.
\item With complement of NP.
\begin{verbatim}
(S (NP-SBJ he)
(VP wrote
(SBAR that
(S (NP-SBJ he)
(VP had
(VP given (PRT up)
(NP hope
(SBAR 0
(S (NP-SBJ they)
(VP would
(ADVP-TMP ever)
(VP agree
(PP-CLR on
(NP anything)))))))))))))
\end{verbatim}
\begin{verbatim}
(S (PP in
(NP the event
(SBAR 0
(S (NP-SBJ Congress)
(VP does
(VP provide
(NP (NP this increase)
(PP in (NP federal funds)))))))))
,
(NP-SBJ the State Board)
(VP should ...))
\end{verbatim}
\end{itemize}
%B.
\subsubsection{Zero relative clauses}
\begin{itemize}
% 1.
\item The null complementizer is labeled (WHNP 0) if it corresponds to
{\it who, which, that,}\/ etc.
\begin{verbatim}
(NP (NP the bird)
(SBAR (WHNP-1 0)
(S (NP-SBJ I)
(VP saw
(NP *T*-1)))))
\end{verbatim}
% 2.
\item The null complementizer is labeled (WHADVP 0) if it corresponds to
{\it where, why, when, how,}\/ etc.
\begin{verbatim}
(NP (NP the place)
(SBAR (WHADVP-2 0)
(S (NP-SBJ I)
(VP put
(NP the book)
(ADVP-PUT *T*-2)))))
\end{verbatim}
\begin{verbatim}
(NP (NP the reason)
(SBAR (WHADVP-3 0)
(S (NP-SBJ he)
(VP came
(ADVP-PRP *T*-3)))))
\end{verbatim}
\end{itemize}
%C.
\subsubsection{Infinitival relative clauses}
See \RefInf\ for more details on Infinitives.
\begin{itemize}
% 1.
\item (WHNP 0) is used for NP objects and subjects.
\begin{verbatim}
(NP (NP a movie)
(SBAR (WHNP-1 0)
(S (NP-SBJ *)
(VP to
(VP see
(NP *T*-1))))))
\end{verbatim}
% 2.
\item (WHADVP 0) is used in the case where the missing element can be
paraphrased as {\it in which, at which, for which,}\/ etc.
\begin{verbatim}
(S (NP-SBJ That)
(VP 's
(NP-PRD (NP a good way)
(SBAR (WHADVP-1 0)
(S (NP-SBJ *)
(VP to
(VP keep
(ADJP-PRD warm)
(ADVP-MNR *T*-1))))))))
\end{verbatim}
\begin{verbatim}
(S (NP-SBJ There)
(VP 's
(NP-PRD (NP no reason)
(SBAR (WHADVP-1 0)
(S (NP-SBJ *)
(VP to
(VP do
(NP it)
(NP-MNR that way)
(ADVP-PRP *T*-1))))))))
\end{verbatim}
% 3.
\item When the infinitival is introduced by {\it for}\/, the appropriate
form of (WHxx 0) is inserted {\bf before}\/ the {\it for}\/ in SBAR:
\begin{verbatim}
(NP (NP a movie)
(SBAR (WHNP-3 0)
for
(S (NP-SBJ us)
(VP to
(VP see
(NP *T*-3))))))
\end{verbatim}
\begin{verbatim}
(NP (NP a good way)
(SBAR (WHADVP-4 0)
for
(S (NP-SBJ them)
(VP to
(VP do
(NP it)
(ADVP-MNR *T*-4))))))
\end{verbatim}
\end{itemize}
\index{zero@0 (null complementizer)|)}
%V.
\subsection{*U* (unit)} \index{U@*U* (unit)|(}
This element marks the interpreted position of a unit symbol, such as \$,
\# (British pounds), FFr (French francs), C\$, US\$, HK\$, A\$, M\$, S\$,
and NZ\$. It may also appear after \% or even {\it cents}\/, when
convenient. See \RefNPmod\ for more details on the use of *U*.
\subsubsection{Uses}
\begin{verbatim}
(NP C$ 5 *U*)
(NP (QP between $ 5 and $ 15) *U*)
\end{verbatim}
After \%, *U* is used only as necessary.
\begin{verbatim}
(NP (QP between 12 % to 13 %)
*U*)
\end{verbatim}
After {\it cents}\/, *U* has been used occasionally for certain complex
constructions, though such use is not officially sanctioned.
\begin{verbatim}
(NP (NP earnings)
(PP of
(NP (NP (QP between 62 cents and 64 cents)
*U*)
(NP-ADV a share))))
\end{verbatim}
\subsubsection{Placement}
In general, *U* is placed where the word corresponding to the symbol would
appear in the string if the text were read aloud. One notable exception is
in certain hyphenated compound adjectives, such as {\it a \$5-a-share
increase}\/ (spoken: ``A five dollar a share increase''). Here, the
bracketing will usually not reflect the spoken order, with *U* placed as
the last element in the ADJP:
\begin{verbatim}
(NP a (ADJP $ 5-a-share *U*)
increase)
\end{verbatim}
Sometimes, this type may lack the *U* entirely.
\index{U@*U* (unit)|)}
%VI.
\subsection{*?* (placeholder for ellipsed material)} \index{*?*|(}
*?* is now available in the following great-tasting flavors: (VP *?*),
(ADJP-PRD *?*), (PP-PRD *), (NP *?*), (S *?*), (SBAR *?*). These act as
placeholders for a missing predicate or piece thereof, especially in
comparative constructions and other environments where predicate deletion
occurs. Although the missing material represented by *?* is often
identical to another constituent in the same sentence, the two are never
coindexed. Postmodifiers of the verb (including traces) may be attached
under (VP *?*), but not to any other null element, including the other *?*
null elements and (VP *T*).
Note that policy for *?* was never finalized, so its use varies to some
extent. In general, *?* is used by the annotators as a last resort (short
of the FRAG analysis) for the annotation of clauses with ``missing''
material. Nonetheless, there are certain constructions that are
particularly likely to contain *?*:
%A.
\subsubsection{Comparative deletion}
(See \RefCompar\ for more information.)
% 1.
\paragraph {Complement of {\it be}\/.} Missing complements of {\it be}\/
are labeled as appropriate and receive the -PRD tag.
\begin{verbatim}
(S (NP-SBJ John)
(VP is
(ADJP-PRD (ADJP sillier)
(SBAR than
(S (NP-SBJ I)
(VP am
(ADJP-PRD *?*)))))))
\end{verbatim}
\begin{verbatim}
(S (NP-SBJ Laos)
(VP is
(PP-PRD of
(NP (NP (ADVP no more)
(ADJP purely military)
value)
(SBAR *ICH*-2)))
(PP to
(NP (NP Moscow)
(NP itself)))
(SBAR-2 than
(S (NP-SBJ it)
(VP is
(PP-PRD *?*)
(PP to
(NP Washington)))))))
\end{verbatim}
% 2.
\paragraph {Direct object.}
\begin{verbatim}
(S (NP-SBJ the Controller)
(VP will
(VP have
(NP (NP the opportunity)
(PP for
(NP (NP greater usefulness)
(PP to
(NP good government))
(SBAR than
(S (NP-SBJ he)
(VP has
(NP *?*)
(ADVP-TMP now))))))))))
\end{verbatim}
\begin{verbatim}
(S (NP-SBJ-2 the Fed)
(VP was
(VP prepared
(S (NP-SBJ *-2)
(VP to
(VP provide
(NP (NP (ADJP as much) credit)
(SBAR as
(S (NP-SBJ the markets)
(VP needed
(NP *?*)))))))))))
\end{verbatim}
% 3.
\paragraph {Verb phrase.}
\begin{verbatim}
(S (NP-SBJ (NP The submission)
(PP of (NP detailed plans)))
(VP would
(VP place
(NP the issues)
(PP-LOC-CLR before (NP the court))
(ADVP-MNR (ADVP more readily)
(SBAR than
(SINV would
(NP-SBJ (NP discussion)
(PP of
(NP divestiture
or
disenfranchisement))
(PP in (NP the abstract)))
(VP *?*)))))))
\end{verbatim}
%should ``readily'' get -MNR??? -m.
% I think so, so I am adding it (to the higher ADVP--didja mean to the
% lower?). -k.
% Hmm. I dunno, but I'll leave it there, I guess.-R.
% 4.
\paragraph {Clausal complement.}
\begin{verbatim}
(S (PP As
(NP Mayor))
,
(NP-SBJ-1 Mr. Levitt)
(VP might
(VP turn
(PRT out)
(S (NP-SBJ *-1)
(VP to
(VP be
(ADJP-PRD (ADJP more independent)
(SBAR than
(S (NP-SBJ (NP some)
(PP of
(NP his leading
supporters)))
(VP would
(VP like
(S *?*))))))))))))
\end{verbatim}
\begin{verbatim}
(S (NP-SBJ the steel strike)
(VP lasted
(ADVP-TMP (ADVP much longer)
(SBAR than
(S (NP-SBJ he)
(VP anticipated
(SBAR 0
(S *?*))))))))
\end{verbatim}
% 5.
\paragraph {VP pro-form {\it do}\/.}
In constructions with a VP pro-form, a missing VP may be postulated and
shown as the complement of {\it do}\/.
\begin{verbatim}
(S (NP-SBJ Bill)
(VP eats
(NP (NP more hotdogs)
(SBAR than
(S (NP-SBJ Mary)
(VP does
(VP *?*)))))))
\end{verbatim}
%B.
\subsubsection{Deletion in non-comparatives}
% 1.
\paragraph {VP after missing auxiliary in second conjunct.}
\begin{verbatim}
(S (S (NP-SBJ She)
(ADVP-TMP rarely)
(VP sings))
,
so
(S (NP-SBJ I)
(VP do n't
(VP think
(SBAR 0
(S (NP-SBJ she)
(VP will
(VP *?*
(NP-TMP tonight)))))))))
\end{verbatim}
\begin{verbatim}
(S (S (NP-SBJ Robin)
(VP likes
(NP ice cream)))
, and
(S (NP-SBJ Kim)
(VP does
(VP *?*
,
(ADVP too)))))
\end{verbatim}
% 2.
\paragraph{VP after missing auxiliary in subordinate clauses.}
\begin{verbatim}
(S (NP-SBJ *)
(VP Call
(S (NP-SBJ it)
(ADJP-PRD anecdotal))
(SBAR-ADV if
(S (NP-SBJ you)
(VP will
(VP *?*))))))
\end{verbatim}
\begin{verbatim}
(S (SBAR-ADV As
(S (NP-SBJ they)
(VP did
(VP *?*
(SBAR-TMP (WHADVP-1 when)
(S (NP-SBJ the Philippines)
(VP was
(NP-PRD a colony)
(ADVP-TMP *T*-1))))))))
,
(NP-SBJ teachers)
(PP for
(NP the most part))
(VP teach
(PP-MNR in
(NP English))))
\end{verbatim}
\begin{verbatim}
(S (NP-SBJ they)
(VP must
(VP buy
(NP shares)
(PP-CLR from
(NP sellers))
(SBAR-TMP (WHADVP-1 when)
(S (NP-SBJ-2 (NP no one)
(ADJP else))
(VP is
(ADJP-PRD willing
(S (NP-SBJ *-2)
(VP to
(VP *?*))))))))))
\end{verbatim}
% 3.
\paragraph{VP missing in {\it as do}\/-type constructions}
\begin{verbatim}
(S (NP-SBJ (NP Warner)
and
(NP Mr. Azoff))
(VP declined
(NP comment)
,
(SBAR-ADV as
(SINV did
(NP-SBJ MCA)
(VP *?*)))))
\end{verbatim}
Non-inverted version ({\it as MCA did}\/) might also have (VP *?*).
\begin{verbatim}
(SBAR-ADV as
(S (NP-SBJ MCA)
(VP did
(VP *?*))))
\end{verbatim}
But note that when the {\it so}\/ predicate pro-form is present, *?* is not
used, and the {\it so}\/ is treated as an adverbial predicate standing in
for the understood verbal predicate:
\begin{verbatim}
(S (NP-SBJ They)
(ADVP also)
(VP did
(ADVP-PRD so)))
\end{verbatim}
\begin{verbatim}
(S (S (NP The winners)
(VP had
(NP fun)))
,
and
(SINV (ADVP-PRD-TPC-1 so)
(VP did
(ADVP-PRD *T*-1))
(NP-SBJ the losers)))
\end{verbatim}
Note, however, that this policy was late in appearance and not always well
understood, so the -PRD label may be missing, or replaced by -CLR or -MNR.
Also, in inverted sentences, the (ADVP so) may lack the -TPC and
accompanying *T*.
% 4.
\paragraph{With relation to previous sentence.}
The following two sentences appear in succession (in wsj\_2106). The second
conjunct of the first sentence contains *?*, indicating ellipsis with
respect to material contained in the first conjunct. The second sentence
also contains *?*, again indicating ellipsis with respect to the same
material. As a rule, we do not indicate intersentential relationships, but
here the second instance of *?* is present by virtue of this
intersentential relationship, though such relationship is not explicitly
recoverable from the annotation.
\begin{verbatim}
( (S Either
(S (NP-SBJ one)
(VP likes
(NP it)))
or
(S (NP-SBJ one)
(VP does n't
(VP *?*)))
.))
( (S (NP-SBJ The typical Glass audience)
(ADVP certainly)
(VP does
(VP *?*))
.))
\end{verbatim}
% If wea are saying that the above indication of intersentential
% relationship is an exception, we should put the rule first, and
% then the exception.
% 5.
\paragraph{In subject position.}
In very rare (about 5) cases, *?* may also appear
in subject position, although * is much more likely to be found there.
\begin{verbatim}
(S (NP-SBJ We)
(VP are
(VP working
(ADVP (ADVP significantly longer and harder)
(SBAR than
(S (NP-SBJ *?*)
(VP has
(VP been
(NP-PRD the case)
(PP-TMP in
(NP the past))))))))))
\end{verbatim}
%Wow! -m.
\label{list:nulls:AsThan-alternatives}
Other possibilities include FRAG:
\begin{verbatim}
(S (NP-SBJ-1 we)
(VP will
(VP take
(NP ongoing cost-reduction actions)
(SBAR-ADV as
(FRAG (ADJP necessary))))))
\end{verbatim}
and (most common by far) simply calling the {\it as}\/ or {\it than}\/ PP:
\begin{verbatim}
(NP-PRD (NP another day)
(PP of
(NP (NP ectoplasmic business)
(PP as
(ADJP usual)))))
\end{verbatim}
\begin{verbatim}
(S (NP-SBJ The decline)
(VP was
(ADJP-PRD (ADJP even steeper)
(PP than
(PP-TMP in
(NP September))))))
\end{verbatim}
%C.
\subsubsection{Undefined gaps}
*?* is occasionally used to fill a (noun-phrase) gap for which there is no
well-established policy:
\begin{verbatim}
(S (NP-SBJ The plant)
(VP will
(VP cost
(NP (QP about 50 million) Canadian dollars)
(S-CLR (NP-SBJ *)
(VP to
(VP build
(NP *?*)))))))
\end{verbatim}
\begin{verbatim}
(NP (NP a return)
(ADJP worth
(S (NP-SBJ *)
(VP getting
(ADJP-PRD excited
(PP about
(NP *?*)))))))
\end{verbatim}
\index{*?*|)}
%VII.
\subsection{*NOT* (anti-placeholder in template gapping)}
\index{NOT@*NOT*|(} \index{gapping!NOT@*NOT*|(} \index{gapping!template}
*NOT* is used in the template gapping procedure, along with ``='' and a
system of coindexation. See \RefCoord\ for more details on the
template gapping approach.
Unlike other null elements, correspondence to a *NOT* is shown by an ``=''
index on the bracket {\bf label}\/, rather than by a ``--'' index on the
null element itself.
*NOT* is used very rarely (about 20 times in the entire WSJ corpus) when
the ``template'' and ``copy'' are not entirely parallel. In principle, it can
serve in two roles:
% maybe these two sections should be reversed -- so the one that appears is
% first and the hypothetical one is second. -a
% Sure, but the second one is so much longer that I think it's clearer this
% way.-R.
% A.
\subsubsection{In the ``copy''}
In the {\bf copy}\/, *NOT* can be used to indicate that the corresponding
constituent in the template is {\bf not}\/ interpreted in the ``copy''.
\begin{verbatim}
(S (NP-SBJ the auditor)
(VP (VP (ADVP-TMP-2 first)
described
(NP-3 the old plan)
(PP-CLR-4 as
(ADJP ill conceived))
(PRN -LRB-
(SBAR-ADV-5 as
(S (NP-SBJ everyone)
(ADVP-TMP already)
(VP agreed)))
-RRB-))
but
(VP (ADVP-TMP=2 then)
(NP=3 (NP the new plan)
(SBAR (WHNP-1 that)
(S (NP-SBJ we)
(VP 'd
(VP worked
(ADVP-MNR so hard)
(PP-CLR on
(NP *T*-1)))))))
(PP-CLR=4 as
(UCP (PP out (PP to (NP lunch)))
and
(ADJP totally half-baked)))
(ADVP=5 *NOT*))))
\end{verbatim}
Note that it is very difficult to construct a grammatical example, so
(unsurprisingly) there are no actual examples of this construction in
the WSJ corpus.
% B.
\subsubsection{In the ``template''}
In the {\bf template}\/, *NOT* appears in the position where the
corresponding constituent in the copy is interpreted, when there is no
matching constituent already in the template.
\begin{verbatim}
(S (NP-SBJ (NP The 189 Democrats)
(SBAR (WHNP-1 who)
(S (NP-SBJ *T*-1)
(VP supported
(NP the override)
(NP-TMP yesterday)))))
(VP compare
(PP-CLR with
(NP (NP (NP-3 175)
(SBAR (WHNP-2 who)
(S (NP-SBJ *T*-2)
(ADVP initially)
(VP backed
(NP the rape-and-incest exemption)
(ADVP-TMP-4 (NP two weeks)
ago)
(PP-5 *NOT*)))))
and
(NP (NP=3 136)
(NP-TMP=4 last year)
(PP=5 on
(NP a similar vote)))))))
\end{verbatim}
More often than not, however, it is assumed that an unindexed constituent
at top level of the ``copy'' is interpreted at highest possible level --
usually VP-level of the template. (Note that this rule doesn't work in the
above example, so it actually needs the *NOT*.)
So the following examples:
\begin{verbatim}
(S (NP The teacher)
(VP (VP gave
(NP-1 Ignatius)
(NP-2 only a B)
(SBAR-3 *NOT*))
,
but
(VP (NP=1 Bertha)
(NP=2 an A)
(SBAR-PRP=3 because
(S (NP-SBJ she)
(VP wrote
(ADVP-MNR so well)))))))
\end{verbatim}
\begin{verbatim}
(VP (VP increasing
(PP-DIR-2 to
(NP 2.5 %))
(PP-TMP-3 in
(NP February 1991))
(ADVP-TMP-4 *NOT*))
,
and
(VP (PP-DIR=2 to
(NP 3 %))
(PP-TMP=3 at
(NP six-month intervals))
(ADVP-TMP=4 thereafter)))
\end{verbatim}
might more likely be bracketed:
\begin{verbatim}
(S (NP The teacher)
(VP (VP gave
(NP-1 Ignatius)
(NP-2 only a B))
,
but
(VP (NP=1 Bertha)
(NP=2 an A)
(SBAR-PRP because
(S (NP-SBJ she)
(VP wrote
(ADVP-MNR so well)))))))
\end{verbatim}
\begin{verbatim}
(VP (VP increasing
(PP-DIR-2 to
(NP 2.5 %))
(PP-TMP-3 in
(NP February 1991)))
,
and
(VP (PP-DIR=2 to
(NP 3 %))
(PP-TMP=3 at
(NP six-month intervals))
(ADVP-TMP thereafter)))
\end{verbatim}
%C.
\subsubsection{Alternatives to *NOT*}
Certain constructions that are sometimes analyzed using *NOT* are more
likely to be analyzed more simply, usually using PRN or FRAG or just PP.
\begin{itemize}
\item complicated *NOT* analysis
\begin{verbatim}
(S (S (NP-SBJ This gap)
(ADVP-TMP-2 eventually)
(VP closes
(ADVP-MNR-1 *NOT*)))
,
but
(S (ADVP-TMP=2 *NOT*)
(ADVP-MNR=1 slowly)))
\end{verbatim}
\item PP analysis
\begin{verbatim}
(S (NP-SBJ This gap)
(ADVP-TMP eventually)
(VP closes
,
(PP but
(ADVP-MNR slowly))))
\end{verbatim}
\item PRN analysis
\begin{verbatim}
(S (NP-SBJ This gap)
(ADVP-TMP eventually)
(VP closes
(PRN ,
but
(ADVP-MNR slowly))))
\end{verbatim}
\item FRAG analysis
\begin{verbatim}
(S (S (NP-SBJ This gap)
(ADVP-TMP eventually)
(VP closes))
,
but
(FRAG (ADVP-MNR slowly)))
\end{verbatim}
\item conjunction? what conjunction?
\begin{verbatim}
(S (NP-SBJ This gap)
(ADVP-TMP eventually)
(VP closes
,
but
(ADVP-MNR slowly)))
\end{verbatim}
%hee hee hee!
\end{itemize}
\index{NOT@*NOT*|)} \index{gapping!NOT@*NOT*|)}
%VIII.
\subsection{Miscellaneous}
%A.
\subsubsection{Subject-aux inversion with subject extractions}
% I s'pose this section should go in separate Wh list. -k.
% Maybe someday, but not today.-R.
When a question is of the subject, the placement of the null *T* subject
determines whether the resulting clause appears to exhibit subject-aux
inversion.
In most cases, the (NP-SBJ *T*) is inserted at the beginning of the SQ,
before the inflected verb or auxiliary:
\begin{verbatim}
(SBARQ (WHNP-4 Who)
(SQ (NP-SBJ *T*-4)
(VP came
(PP-DIR to
(NP the party))))
?)
\end{verbatim}
\begin{verbatim}
(SBARQ (WHNP-1 Who)
(SQ (NP-SBJ *T*-1)
(VP will
(VP come
(PP-DIR to
(NP the party))))))
\end{verbatim}
\begin{verbatim}
(SBARQ (WHNP-4 Who)
(SQ (NP-SBJ *T*-4)
(ADVP-TMP always)
(VP comes
(PP-DIR to
(NP parties))))
?)
\end{verbatim}
\begin{verbatim}
(SBARQ (WHNP-1 Who)
(SQ (NP-SBJ *T*-1)
(VP will
(ADVP-TMP never)
(VP come
(PP-DIR to
(NP the party))))))
\end{verbatim}
Note, however, that there is some variation (since annotators are
accustomed to seeing subject-aux inversion in many questions), and the
subject trace is occasionally inserted directly after an initial
auxiliary:
\begin{verbatim}
(SBARQ (WHNP-1 Who)
(SQ will
(NP-SBJ *T*-1)
(VP come
(PP-DIR to
(NP the party)))))
\end{verbatim}
\begin{verbatim}
(SBARQ (WHNP-1 Who)
(SQ will
(NP-SBJ *T*-1)
(ADVP-TMP never)
(VP come
(PP-DIR to
(NP the party)))))
\end{verbatim}
%B.
\subsubsection{Reduced relatives} \index{relative clauses!reduced
relatives!w/subj-aux inversion}
\label{list:nulls:RRCs}
Reduced relative clauses are bracketed as follows,
with no structure above the VP level:
\begin{verbatim}
(NP (NP an elephant)
(VP called
(S (NP-SBJ *)
(NP-PRD Dumbo))))
\end{verbatim}
It may be that the underlying structure of the reduced relative is as
follows:
\begin{verbatim}
(NP (NP an elephant)
(SBAR (WHNP-1 0)
(S (NP-SBJ-2 *T*-1)
(VP BE
(VP called
(S (NP-SBJ *-2)
(NP-PRD Dumbo)))))))
\end{verbatim}
But we do not attempt to reflect this particular understanding of the
reduced relative in the annotation. Nonetheless, the bracketing as it is
can automatically be transformed into such a structure. Note also that
choosing the former style of annotation over the latter necessarily affects
coindexation: The passive (NP *) does not bear an index in the former, as
the null subject with which it is associated is not present in the
annotation.
%C.
\subsubsection{Attachment to null elements} \index{null elements!attaching to}
Barring certain exceptions, a null element never has another constituent
attached to it. Thus, the SBAR in the first example below and the PP in
the second example are not attached to the trace, but rather to its overt
associate NP elsewhere in the sentence.
\begin{verbatim}
(SBARQ (WHNP-2 (WHNP Who)
(SBAR *ICH*-3))
(SQ did
(NP-SBJ you)
(VP meet
(NP *T*-2)
(SBAR-3 (WHNP-1 that)
(S (NP-SBJ *T*-1)
(VP wore
(NP overalls)))))))
\end{verbatim}
\begin{verbatim}
(SINV (ADVP-LOC-PRD-TPC-5 Here)
(VP are
(ADVP-LOC-PRD *T*-5))
(NP-SBJ (NP the pictures)
(SBAR (WHNP-3 0)
(S (NP-SBJ-1 you)
(VP wanted
(S (NP-SBJ *-1)
(VP to
(VP see
(NP *T*-3)))))))
(PP of
(NP (NP that cute dog)
(SBAR (WHNP-2 0)
(S (NP-SBJ we)
(VP met
(NP *T*-2)
(NP-TMP the other day))))))))
\end{verbatim}
%oh, my, god.
Note that (VP *?*) (but {\bf not}\/ (VP *T*-1)) is an exception to this rule.
\begin{verbatim}
(S (S (NP-SBJ She)
(ADVP-TMP rarely)
(VP sings))
,
so
(S (NP-SBJ I)
(VP do n't
(VP think
(SBAR 0
(S (NP-SBJ she)
(VP will
(VP *?*
(NP-TMP tonight)))))))))
\end{verbatim}
%if we officially do not treat ``so'' as a conjunction, should we change
%this one to ``and'' or subordinate it, or some such? -a
%Certainly not subordinate. My guess is that what's above is what we would
%most likely do with it, so it might as well stay as is.-R.
%D.
\subsubsection{Attachment of null elements}
%1.
\paragraph{*PPA* (permanent predictable ambiguity).}
*PPA*-attach is used to
indicate ambiguity of attachment of a trace, if the sentence is truly
ambiguous (here ``why was the decision made'' vs. ``why do you think it was
made''):
\begin{verbatim}
(SBARQ (WHADVP-1 Why)
(SQ do
(NP-SBJ you)
(VP think
(SBAR 0
(S (NP-SBJ we)
(VP made
(NP that decision)
(ADVP-PRP *PPA*-2))))
(ADVP-PRP-2 *T*-1))))
\end{verbatim}
Note that such ambiguity is unlikely context, so such examples are rare or
nonexistent in the actual corpus.
%2.
\paragraph{Shared traces.} \index{shared traces}
On the other hand, {\bf shared}\/ traces are handled
quite differently. When a trace is interpreted as part of two separate
conjuncts, there will be one trace at conjuction level if the element in
question is a VP adjunct:
\begin{verbatim}
(NP (NP a business system)
(SBAR (WHADVP-1 where)
(S (NP-SBJ shareholders)
(VP (VP have
(NP few rights))
and
(VP expect
(NP only modest dividends))
(ADVP-LOC *T*-1)))))
\end{verbatim}
...but two separate traces otherwise (pseudo-attach is not used):
\begin{verbatim}
(S (NP-SBJ-2 (NP (QP No fewer than 24)
country funds))
(VP have
(VP been
(VP (VP launched
(NP *-2))
or
(VP registered
(NP *-2)
(PP-CLR with
(NP regulators)))
(NP-TMP this year)))))
\end{verbatim}
\begin{verbatim}
(S (PP-TPC-1 Of (NP the 13 entrants))
,
(S (NP-SBJ (NP 5)
(PP *T*-1))
(VP finished))
and
(S (NP-SBJ (NP 8)
(PP *T*-1))
(VP crashed)))
\end{verbatim}
%E.
\subsubsection{Interpreting the WH label} \index{wh-clauses@{\it wh}\/-clauses}
% Some of this is probably covered in the Wh list, but a little redundancy
% won't kill us.-R.
WHNP, WHADJP, etc. are labels that mark a \iwh-phrase in SBAR that
has an associated trace *T* in the position where the \iwh-phrase is
interpreted.
{\it Wh}\/-phrases usually contain a \iwh-word, such as {\it who,
whose, which, when, where, how, why, whom, whenever, whatever,}\/ etc. in
questions and relative clauses. The WHx label is also used for {\it
that}\/ and 0 (zero)\/ in relative clauses.
The label applies only to \iwh-words that appear in SBAR. It is not
used for {\it in situ} \iwh-phrases:
\begin{verbatim}
(S (NP-SBJ You)
(VP said
(NP what))
?)
\end{verbatim}
\begin{verbatim}
(S (NP-SBJ The butcher)
(VP gave
(NP the bone)
(PP-DTV to
(NP which dog)))
?)
\end{verbatim}
However, note that in sentence fragments where the relative position of the
\iwh-word may not be clear, the \iwh-word in question may be
labeled as such if the annotator has the sense that the wh word is not {\it
in situ}.
\begin{verbatim}
(FRAG (WHADVP Why)
not
?)
\end{verbatim}
\begin{verbatim}
( (S (NP-SBJ You)
(VP said
(SBAR 0
(S (NP-SBJ John)
(VP gave
(NP you)
(NP something)))))
.))
( (FRAG (WHNP what)
?))
\end{verbatim}
In complex \iwh-phrases, \iwh-ness percolates up but not down.
Thus, in the following, {\it from}\/ is labeled WH but {\it of}\/ and {\it
syntax}\/ are not.
\begin{verbatim}
(SBARQ (WHPP-1 From
(WHNP (WHNP whose theory)
(PP of
(NP syntax))))
(SQ do
(NP-SBJ you)
(VP draw
(NP this conclusion)
(PP-CLR *T*-1)))
?)
\end{verbatim}
%F.
\subsubsection{Comparative relatives} \index{relative clauses!comparative}
Certain comparatives may have been analyzed as relative clauses, as in (b),
although the *?* analysis is more common, as in (a):
\begin{verbatim}
a. (S (NP-SBJ-2 the Fed)
(VP was
(ADJP-PRD prepared
(S (NP-SBJ *-2)
(VP to
(VP provide
(NP (NP (ADJP as much) credit)
(SBAR as
(S (NP-SBJ the markets)
(VP needed
(NP *?*)))))))))))
\end{verbatim}
\begin{verbatim}
b. (S (NP-SBJ-2 the Fed)
(VP was
(ADJP-PRD prepared
(S (NP-SBJ *-2)
(VP to
(VP provide
(NP (NP (ADJP as much) credit)
(SBAR (WHNP-3 0)
as
(S (NP-SBJ the markets)
(VP needed
(NP *T*-3)))))))))))
\end{verbatim}
%G.
\subsubsection{Illegal null elements} \index{null elements!illegal types}
\label{list:nulls:illegal}
Following is a list of old and/or improperly formed null elements. They
should be removed or updated as described, but they may occasionally slip
into published files, despite checks which are designed to prevent this.
\begin{codelist}
\item[*T*] should always have a reference index. Any *T* that lacks a
reference index should be indexed as appropriate, or removed.
\index{T@*T*!must have reference index}
\item[*pseudo-attach*] used to be an all-purpose pseudo-attach marker. This
should be converted to *ICH*, *RNR*, *EXP*, or *PPA*, as appropriate, or
removed if now unnecessary. Note that the reference index for this null
element appeared on the bracket label, making it appear to be an identity
index. \index{pseudo-attach@*pseudo-attach*}
\item[T] formerly stood for noun phrase \iwh-traces. It should be
changed to *T* and indexed as appropriate. Note, however, that it is
possible for `T' to appear as part of real text (for example, as a symbol
for `Temperature' in scientific writing). \index{T}
\item[+] is inserted by the {\sc Fidditch}\/ parser as a passive trace.
These are fixed in preprocessing, but those that slip through should be
removed or replaced with `*'. \index{+}
\item[OF] is inserted by the {\sc Fidditch}\/ parser after predeterminers
such as {\it all}\/. It should be removed, along with its accompanying PP,
and the noun phrases should be flattened into a single NP: \index{OF}
\begin{verbatim}
(NP (NP all)
(PP OF
(NP the dogs)))
\end{verbatim}
should be changed to:
\begin{verbatim}
(NP all the dogs)
\end{verbatim}
\end{codelist}
Note also that it is possible for a null element to still have a reference
index after the brackets with the corresponding identity index have been
removed. However, such errors are relatively rare (about 0.1\% of all
indexed null elements).
%H.
\subsubsection{Limits of coindexation} \index{coindexing!limits of}
\label{list:nulls:coindex-limits}
The \code{*} null element generally receives a reference index whenever
there is an appropriate referent elsewhere in the same sentence. However,
there are cases in which annotators tend to not coindex, even when they can
find a plausible referent. Some of these criteria overlap with each other
and with rules described above.
%1.
\paragraph{Non-arguments.}
Annotators usually avoid indexing from non-arguments.
\begin{verbatim}
(S (PP for
(NP us))
,
(S-NOM-SBJ (NP-SBJ *)
(VP eating
(NP chocolate)))
(VP is
(NP-PRD (NP a way)
(PP of
(NP life)))))
\end{verbatim}
\begin{verbatim}
(S (PP For
(NP Willie))
,
(NP-SBJ (NP it)
(S *EXP*-1))
(VP is
(ADJP-PRD difficult)
(S-1 (NP-SBJ *)
(VP to
(VP resist
(NP chocolate))))))
\end{verbatim}
\begin{verbatim}
(S (PP for (NP Zaphod))
,
(NP-SBJ that steak)
(VP is
(ADJP-PRD ready
(SBAR (WHNP-1 0)
(S (NP-SBJ *)
(VP to
(VP eat
(NP *T*-1))))))))
\end{verbatim}
%2.
\paragraph{Gerund PP objects.}
Null subjects of gerund complements of PP modifiers of NPs are coindexed
only if there is a particularly strong coindexed interpretation or the PP
appears to be part of some ``fixed phrase''.
%overall c. 50% of the relevant cases have a coindexed interpretation,
%however, only c. 10% overall are bracketed with coindexation.
\begin{verbatim}
(S (NP-SBJ the company)
(VP has
(NP (NP no intention)
(PP of
(S-NOM (NP-SBJ *)
(VP tapping
(NP its short-term bank lines)))))
(PP-TMP for
(NP (NP a good part)
(PP of
(NP 1990))))))
\end{verbatim}
\begin{verbatim}
(S (PP In
(NP (NP addition)
(PP to
(S-NOM (NP-SBJ *-1)
(VP having
(NP high price-earnings ratios))))))
,
(NP-SBJ-1 most)
(VP pay
(NP puny dividends)))
\end{verbatim}
%3.
\paragraph{Possessive NPs.}
NP brackets that only mark a possessive phrase within an NP should NOT
serve as a referent for a * null element:
% from wsj_0961
\begin{verbatim}
(S (NP-SBJ (NP the Fed 's)
goal)
(VP is
(S-PRD (NP-SBJ *)
(VP to
(VP reduce
(NP inflation))))))
\end{verbatim}
% from wsj_2384
\begin{verbatim}
(NP (NP (NP Mr. Bush 's)
claim
(SBAR that...))
and
(NP (NP his insistence)
(PP on
(S-NOM (NP-SBJ *)
(VP combining...)))))
\end{verbatim}
However, a possessive NP that is acting as a subject may serve such a role:
%from wsj_0305
\begin{verbatim}
(S (NP-SBJ-1 Pinkerton 's)
(VP had
(VP locked
(NP itself)
(PP-CLR into
(NP low-price contracts))
(S-PRP (NP-SBJ *-1)
(VP to
(VP win
(NP new business)))))))
\end{verbatim}
%4.
\paragraph{Indexing in modified NPs.}
When an NP is adjoined with modifiers, the head NP should not serve as a
referent for a * null element (although it may be used for template
gapping). The adjunction level should also not serve as a reference for
*'s found within the NP's modifiers.
% tgrep '/^NP/ < /^NP.*-[0-9]*$/' gets some exceptions, some of which are
% errors and some of which are gapping templates
In the following example, the * subject of {\it spotting}\/ should not be
indexed either to {\it New York money manager Mario Gabelli}\/ or {\it an
expert}\/, nor should it be indexed to the whole NP-SBJ.
%from wsj_2376
\begin{verbatim}
(NP-SBJ (NP New York money manager Mario Gabelli)
,
(NP (NP an expert)
(PP at
(S-NOM (NP-SBJ *)
(VP spotting
(NP takeover candidates)))))
,)
\end{verbatim}
In the following example, the * subject of {\it buy}\/ should not be indexed,
although it is clear that the {\it Soviet companies}\/ are doing the buying.
% from wsj_0934
\begin{verbatim}
(NP-SBJ-2 (NP Soviet companies)
(VP needing
(NP Western currencies)
(S-PRP-CLR (NP-SBJ *)
(VP to
(VP buy
(NP equipment and supplies)
(ADVP-LOC abroad))))))
\end{verbatim}
The following is an error, since the coindexation should have been from the
whole subject noun phrase:
% from wsj_0224
\begin{verbatim}
(S (NP-SBJ (NP-1 Ford)
,
(SBAR (WHNP-64 which)
(S (NP-SBJ *T*-64)
(ADVP-TMP already)
(VP has
(NP an unwelcome (ADJP 13.2 %) holding))))
,)
(VP is
(ADJP-PRD prepared
(S (NP-SBJ *-1)
(VP to
(VP bid
(PP-PRP for
(NP the entire company))))))))
\end{verbatim}
\index{null elements|)}
%\printindex
%\end{document}
diff --git a/layout.md b/layout.md
index 0058b58..e69de29 100644
--- a/layout.md
+++ b/layout.md
...
Id_nulls_tex_v_1__.tex
