<?tei:SOF p2aicceg.odd ?>
<!-- Revisions:                                               -->
<!-- 93-06-03 : MSM : paper over broken taglist               -->
<!-- 93-06-03 : MSM : fix typo                                -->
<!-- 93-06-03 : MSM : cosmetics                               -->
 
<div3 id=AICCEG><head>An Example of Simple Linguistic Analysis</head>
 
<p>In this section we discuss a number of ways in which simple
linguistic analyses may be represented, using as an example, the
following sentence, taken from the Lancaster/IBM Treebank
Project:<note>
See <citation> <author>G.  N. Leech and R. G. Garside,</author>
<title.p>Running a Grammar Factory,</title.p> in
<title>English Computer Corpora:
Selected Papers and Research Guide,</title>
ed. <editor>S. Johansson and A.-B.  Stenstr&ouml;m</editor>
(Berlin: de Gruyter; New York:  Mouton, 1991),
pp. 15-32.
</citation>
This sentence and its analysis are reproduced by kind permission of the
University of Lancaster's Unit for Computer Research on the English
Language.
<!-- Stenstrom's name corrected to ouml because she's Swedish (LB) -->
<!-- Stenstrom's name changed from oelig, on theory that      -->
<!-- he's Scandinavian not French.  Pls check.  -MSM          -->
</note>
 
<!-- [N [G The_AT victim_NN1 's_$ G] friends_NN2 N] [V told_VVD  -->
<!-- [N police_NN2 N] [Fn that_CST [N Krueger_NP1 N] [V [V&   -->
<!-- drove_VVD [P into_II [N the_AT quarry_NN1 N]P]V&]        -->
<!-- and_CC [V+ never_RR surfaced_VVD V+]V]Fn]V] ._.          -->
 
<q rend=display>
The victim's friends told police that Kruger drove into the quarry
and never surfaced.
</q>
 
<p>Our discussion focuses on the way that this sentence might be
analysed using the Claws system developed at the University of
Lancaster, but exactly the same principles may be applied to a wide
variety of other systems.<note>For the word-class tagging method used by
Claws see
<citation>
<author>I. Marshall</author>,
<title.p>Choice of Grammatical Word Class without Global Syntactic
Analysis:  Tagging Words in the LOB Corpus,</title.p>
in
<title>Computers and the Humanities</title>
<detail>17 (1983): 139-50</detail>.
</citation>
 
For an overview of the system see
 
<citation>
<author>R. G. Garside, G. N. Leech, and G. R. Sampson,</author>
<title>The Computational Analysis of English: a Corpus-Based
Approach</title>
(Oxford:  Oxford University Press, 1987).
</citation>
</note>
 
Output from the system consists of a segmented and tokenized version of
the text, in which word class codes have been associated with each
token.  For our example sentence, we might conveniently represent these
codes using entity references:
 
<eg><![ CDATA [
<s>The&AT victim&NN1 's&GEN friends&NN2 told&VVD police&NN2
that&CST Krueger&NP1  drove&VVD into&II the&AT
quarry&NN1  and&CC  never&RR surfaced&VVD
</s>
]]>
</eg>
<!-- added explanatory note on GEN (LB) -->
The names used for these entity references have some significance for
the human reader (<q>AT</q> for article, <q>NN1</q> for singular noun,
<q>NN2</q> for plural noun, etc.), but their representation in the
output from an SGML system processing the document may be chosen to
suit the convenience of whatever analytic software is to be used.<note>
The name <q>gen</q> is used to represent the dollar sign code used by
the CLAWS system to denote a genitive ending</note>  For
example, if the SGML parser operating on this sentence uses a set of
entity declarations in the following form:
 
<eg><![ CDATA [
<!ENTITY AT "">
<!ENTITY NN1 "">
<!-- ... -->
]]>
</eg>
 
then the wordclass tags will simply disappear from the output.
Alternatively, if the entity set in use follows the following pattern:
 
<eg><![ CDATA [
<!ENTITY AT  "[definite article]">
<!ENTITY NN1 "[singular noun]">
<!ENTITY gen "[genitive suffix]">
<!-- ... -->
]]>
</eg>
 
then our sample sentence will be processed by an SGML-aware processor as
if it began
 
<eg><![ CDATA [
The[definite article] victim[singular noun] 's[genitive suffix] ...
]]>
</eg>
<!-- toned down to appease G. Sampson (LB) -->
<p>Alternatively, replacement texts for each entity might be codes
of some significance to a particular analysis program.  For example,
a replacement might be chosen to make explicit
the fact that (for example) <q>NN1</q> and <q>NN2</q> have
something in common, namely their <q>noun-ness</q>, which they do not
share with (say) <q>VVD</q>.  It is a matter for the analyst to decide
whether or not the word class categories represented by these codes are
decomposable into bundles of more primitive analytic features.  In cases
where they are, such analytic features may be defined as a set of
<term>feature structures</term>, as described in chapters <xref
target=FS> and <xref target=FD>.
 
<!-- Presumably held as an < gi >fsLib< / gi > element        -->
<!-- within the < gi >analysis< / gi > element of the         -->
<!-- encoding declarations in the < gi >TEIHeader< / gi >     -->
<!-- but I'm not sure:  need to change to header dtd to       -->
<!-- permit this if so.  (LB)                                 -->
 
A suitable replacement for the word-class entity references above would
then be an empty <gi>anchor</gi> element bearing a <att>struct</att>
attribute (for which see chapter <xref target=FS>).
<!-- Check target:  FS or AI? -MSM                            -->
<!-- Should it be a milestone? or an anchor? or a ptr?  If    -->
<!-- it's a ptr, should the analysis be specified by the      -->
<!-- target attrib?  (LB)                                     -->
 
<eg><![ CDATA [
<s>The <anchor struct=AT> victim <anchor struct=NN1>
's <anchor struct=amp> friends <anchor struct=NN2>
told <anchor struct=VVD> police <anchor struct=NN2>
that <anchor struct=CST> Krueger <anchor struct=NP1>
drove <anchor struct=VVD> into <anchor struct=II>
the <anchor struct=AT> quarry <anchor struct=NN1
and <anchor struct=CC>  never <anchor struct=RR>
surfaced <anchor struct=VVD>
</s>
]]>
</eg>
 
The <gi>anchor</gi> element simply marks a place in the text.  The
<att>struct</att> attribute associates that place with the relevant
feature structure by specifying its unique identifier.  This solution
requires that there be somewhere a formal definition for what exactly is
meant by analytic code <q>NN1</q>, in terms as simple or as complex as
seems appropriate to the analyst.  For example, we might represent the
constituent features of the codes AT and NN1 using feature
structure notation as follows:
 
<eg>
<![ CDATA [
<fs id=AT>
   <f name=formClass><sym value=article>
</fs>
<!-- ... -->
<fs id=NN1>
   <f name=formClass><sym value=noun>
   <f name=proper><minus>
   <f name=number><sym value=singular>
</fs>
]]>
</eg>
 
For a more detailed presentation of this method of representing the
analytic features underlying these word class code, refer to chapter
<xref target=FS>.
 
<p>Although common practice, this method of presenting a token-level
analysis requires application of an unstated convention about how the
input stream is to be tokenized.  The interpretive structures are
associated not with a distinct stretch of text, but with a point in it,
by convention immediately following the token to which they
<soCalled>really</soCalled> apply.  This convention makes it impossible
to associate interpretive structures directly with sequences longer than
a single token.
 
<p>Neither of these inconveniences applies if the text is fully
segmented.
<!--  < taglist >                                             -->
<!--  < tagdesc tagDoc=SEG atts='type' >                      -->
<!--  < tagdesc tagDoc=C atts='type' >                        -->
<!--  < tagdesc tagDoc=W atts='type' >                        -->
<!--  < tagdesc tagDoc=M atts='type' >                        -->
<!--  < /taglist >                                            -->
<!-- taglist was broken, following text is to paper over loss. -->
A variety of special-purpose elements are provided for
segmentation at various levels, including <gi>seg</gi> (generic
segment),
<gi>m</gi> (morpheme),
<gi>w</gi> (word),
<gi>c</gi> (constituent),
<gi>phr</gi> (phrase), and
<gi>cl</gi> (clause).  Each has a <att>type</att> attribute for
further subcategorization.
 
In the following example, the general purpose <gi>seg</gi> element is
used to represent each token with the outer <gi>s</gi> element
containing the whole orthographic sentence:
 
<eg><![ CDATA [
<s>
<seg struct=AT  >The</seg>
<seg struct=NN1 >victim</seg>
<seg struct=amp >'s</seg>
<seg struct=NN2 >friends</seg>
<seg struct=VVD >told</seg>
<seg struct=NN2 >police</seg>
<seg struct=CST >that</seg>
<seg struct=NP1 >Krueger</seg>
<seg struct=VVD >drove</seg>
<seg struct=II  >into</seg>
<seg struct=AT  >the</seg>
<seg struct=NN1 >quarry</seg>
<seg struct=CC  >and</seg>
<seg struct=RR  >never</seg>
<seg struct=VVD >surfaced</seg>
</s>
]]>
</eg>
 
Unlike the <gi>s</gi> element, the <gi>seg</gi> element can nest within
itself, and may thus be used to represent segmentation carried out at
several levels, provided that the whole of the segmentation is
describable by a single hierarchic tree.  For example, it might be used
to mark both constituent phrases, word-like units, and even a
lower-level morphological level of analysis.  To distinguish amongst
these various uses of the <gi>s</gi> element, the <att>type</att>
attribute may be used, as in the following example:
 
 
<eg><![ CDATA [
<s>
   <seg type=phrasal>
      <seg type=lexical struct=AT>The</seg>
      <seg type=lexical struct=NN1>
         <seg type=morph struct=NN1>victim</seg>
         <seg type=morph struct=GEN>'s</seg>
      </seg>
      <seg type=lexical struct=NN2>friends</seg>
   </seg>
   <seg type=phrasal>
      <seg type=lexical struct=VVD>told</seg>
   </seg>
   <seg type=phrasal>
      <seg type=lexical struct=NN2>police</seg>
   </seg>
      <!-- ... -->
</s>
 
]]>
</eg>
 
<p>In projects where annotation is routinely carried out at multiple
levels it will often be felt preferable to use a more specific element
appropriate to each level of analysis.    The <gi>c</gi>, <gi>w</gi>
and <gi>m</gi> elements are provided for  this purpose. Thus, <gi>c</gi>
may be thought of as equivalent to <tag>seg type=constituent</tag>;
<gi>w</gi> as equivalent to <tag>seg type=word</tag> and <gi>m</gi> as
equivalent to <tag>seg type=morph</tag>. Using these,  the start of our
example sentence might be tagged as follows:
<eg>
<![ CDATA [
   <c>
      <w>The</w>
      <w><m>victim</m>
         <m>'s</m>
      </w>
      <w>friends</w>
   </c>
]]>
</eg>
 
As a further example, Claws provides the following constituent analysis
of the sample sentence:
 
<q rend=display>
[N [G The_AT victim_NN1 's_$ G] friends_NN2 N] [V told_VVD
[N police_NN2 N] [Fn that_CST [N Krueger_NP1 N] [V
[V& drove_VVD [P into_II [N the_AT quarry_NN1 N]P]V&]
and_CC [V+ never_RR surfaced_VVD V+]V]Fn]V] ._.
</q>
 
<p>The <gi>c</gi> element may be used to represent this analysis, as
follows:
 
<eg><![ CDATA [
<s>
<c struct=N>
   <c struct=G> The victim's </c>
   friends
</c>
<c struct=V>
   told
   <c struct=N> police </c>
   <c struct=Fn>
      that
      <c struct=N> Krueger </c>
      <c struct=V>
          <c id=C1 struct=VX next=c2> drove
             <c struct=P> into
                <c struct=N> the quarry </c>
             </c>
          </c>
          and
          <c id=C2 struct=VX prev=c1> never surfaced </c>
      </c>
   </c>
</c>
</s>
]]>
</eg>
 
In the above analysis, the value borne by each <att>struct</att>
attribute corresponds with one of a set of feature structures describing
the constitutive function of each sequence of tokens.  For example,
<term>N</term> is a <soCalled>noun phrase</soCalled>, <term>G</term> a
<soCalled>genitive</soCalled>, <term>FN</term> a <soCalled>noun
clause</soCalled>, and <term>V</term> a <soCalled>verb
phrase</soCalled>.  The feature structures indicated will contain
further information about the classification used, in the same way as
for the word-level tagging described above.  Alternatively, if no
further specification for the labels attached to constituents is
required, then a <att>type</att> attribute might be defined for the
<gi>c</gi> element.  In such a case, the significance of the values
specified should be specified in the <gi>encodingDecl</gi> element of
the header.
 
<div3 id=CCAN3><head>Non-nesting Structures</head>
 
<p>Each analytic segment so far discussed has been well-behaved with
respect to the basic document hierarchy, having only a single parent.
Moreover, the segmentation has been complete, in that each part of the
text is accounted for by some segment at each level of analysis, without
discontinuities or overlap.  This happy state of affairs does not of
course apply in all types of analysis, and these Guidelines therefore
provide a number of mechanisms to support the representation of
discontinuities or multiple analyses.  A brief overview of these
facilities is provided in chapter <xref target=NH>.
 
<!-- some specific examples are discussed in this section.    -->
 
<p>These mechanisms all depend to a greater or less degree on the
ability to associate a unique identifier with any element in a
TEI-conformant text, and then to specify that identifier as the target
of a pointing element of some kind.  The previous section included one
example of such a linkage in its use of the <att>next</att> and
<att>prev</att> attributes to represent the fact that the two
<term>conjunct</term> verbal phrases in the example may be aggregated
together.  A similar technique might be used to represent anaphoric
links, for example the fact that <q>the victim</q> and <q>Krueger</q> in
this example both refer to the same person:
 
<eg><![ CDATA[
 
<c struct=N corresp=N1>The victim</c> 's
   friends
<c struct=V>
   told
   <c struct=N> police </c>
   <c struct=Fn>
      that
      <c struct=N id=N1> Krueger </c>
<!-- ... -->
]]>
</eg>
 
These attributes, and a number of other related topics, are further
discussed in section <xref target=SAAG> and <xref target=SACS>.
 
 
<p>The same mechanisms may also be used to encode analyses of an
entirely different kind.  Here, for example, the feature structures
being associated with the various segments of the text are concerned
with their discoursal function (<q>df</q>):
 
<eg>
<![ CDATA [
<seg type=df struct=SR>
   <u who=P1 id=U1>Can I have ten oranges and a kilo
   of bananas please?
</seg><seg type=df struct=SC>
   <u who=P2 id=U2>Yes, anything else?
   <u who=P1 id=U3>No thanks.
</seg><seg type=df struct=S>
   <u who=P2 id=U4>That'll be dollar forty.
</seg><seg type=df struct=P>
   <u who=P1 id=U5>Two dollars
</seg><seg type=df struct=PC>
   <u who=P1 id=U6>Sixty, eighty, two dollars.  Thank you.
</seg>
]]>
</eg><!-- R. Hasan `Language, context, and text... OUP1989, p 59 -->
 
<p>The associated feature structures will document any internal
substructure the analyst wishes to associate with each category assigned
to the various parts of the discourse above, or simply document them
further; in this case specifying that <term>SR</term> is a
<soCalled>sale request</soCalled>, <term>SC</term> a <soCalled>sale
compliance</soCalled>, <term>PC</term> a <soCalled>purchase
closure</soCalled> and so on.
 
<!-- SR = sales request; SC=sale compliance; S=sale;          -->
<!-- P=purchase; PC = purchase closure                        -->
 
For further discussion of the <gi>u</gi> (utterance) element and other
elements recommended for transcriptions of spoken language, see chapter
<xref target=TS>.
 
<!-- more example to come here -->
 
 
<?tei:EOF p2aicceg.odd ?>
