Hypertags 
Alexandra KINYON 
Talana / Lattice, Univ. Paris 7 
UFRL case 7003 
2 pl Jussieu 75005 Paris, France 
Alex andra.Kinyon @ linguist.j ussieu.fr 
Abstract 
Srinivas (97) enriches traditional 
morpho-syntactic POS tagging with 
syntactic information by introducing 
Supertags. Unfortunately, words are 
assigned on average a much higher number 
of Supertags than traditional POS. In this 
paper, we develop the notion of Hypertag, 
first introduced in Kinyon (00a) and in 
Kinyon (00b), which allows to factor the 
information contained in ~everal Supertags 
into a single structure and to encode 
flmctional information in a systematic 
lnanner. We show why other possible 
solutions based on mathematical properties 
of trees are unsatisfactory and also discuss 
the practical usefulness of this approach. 
Introduction 
As a first step prior to parsing, traditional Part 
of Speech (POS) tagging assigns limited 
morpho-syntactic information to lexical items. 
These labels can be more or less fine-grained 
depending on the tagset , but syntactic 
information is often absent or limited. Also, most 
lexical items are assigned several POS. Although 
lexical ambiguities are dealt with by POS taggers, 
either in a rule-based or in probabilistic manner, it 
is useful to delay this decision at a further parsing 
step (e.g. Giguet (98) shows that knowing 
constituent boundaries is crucial for solving 
lexical ambiguity correctly). In order to do so, it 
would help to be able to encode several POS into 
one compact representation. 
In order to assign richer syntactic information 
to lexical items Joshi & Srinivas (94) and 
Srinivas (97) introduce the notion of Supertags, 
developed within the fiamework of Tree 
Adjoining Grammars (TAG). The idea behind 
Supertags is to assign to each word in a sentence, 
instead of a traditional POS, an "elementary 
tree", which constitutes a primitive syntactic 
structure within the TAG frmnework. A 
supertagged text can then be inputed to a parser 
or shallow parser, thus alleviating the task of the 
parser. Several problems remain though: 
• Even when no lexical ambiguity occurs, each 
word can anchor several trees (several hundreds 
for some verbs) I. On average for English a word 
is associated with 1.5 POS and with 9 supertags 
(Joshi (99)). One common solution to the 
problem is to only retain the "best" supertag for 
each word, or eventually the 3 best supertags for 
each word, but then early decision has an adverse 
effect on the quality of parsing if the wrong 
supertag(s) have been kept : one typically obtains 
between 75% and 92% accuracy when 
supertagging, depending on the type of text being 
supertagged and on the technique used) (cf 
Srinivas (97), Chen & al (99), Srinivas & Joshi 
(99)). This means that it may be the case that 
every word in 4 will be assigned the wrong 
supertag, whereas typical POS taggers usually 
achieve an accuracy above 95%. 
• Supertagged texts rely heavily on the TAG 
framework and therefore may be difficult to 
exploit without being familiar with this 
fornaal ism. 
• Supertagged texts are difficult to read and 
thus difficult to annotate manually. 
• Some structural information contained in 
Supertags is redundant 
• Some information is missing, especially with 
respect to syntactic functions 2. 
So our idea is to investigate how supertags can 
be underspecified so that instead of associating a 
set of supertags to each word, one could associate 
one single structure, which we call hypertag, and 
which contains the same information as a set of 
supertags as well as functional information 
Our practical goal is fourfolds : 
a) delaying decision for parsing 
b) obtaining a compact and readable 
representation, which can be manually annotated 
1 See Barrier & al. (00) for precise data for French, using 
the FFAG wide-coverage grammar developped at 
TALANA, University of I'aris 7. 
2 The usefulness of functional information ill POS tagging 
has also been discussed within the reductionist paradigm 
(cf Voutilainen & Tapanainen (93)). 
446 
as a step towards building a treebank for French 
(cf Abeill6 & al. (00a), Cl6ment & Kinyon (00)). 
c) extracting linguistic information on a large 
scale such as lcxical preferences for verb 
subcategorization frames. (cf Kinyon (99a)) 
(1) Building an efficient, but nonetheless 
psycholinguistically motivated, processing model 
for TAGs (cf Kinyon (99b)) 
Ttms, in addition of being well-defined 
computational objects (Point a), hypertags should 
I)e "readable" (point b) and also motivated from a 
linguistic point of view (Points c & d). 
In the first part of this paper, we briefly 
introduce the LTAG frmnework and give 
exmnples of supertags. In a second part, we 
investigate several potential ways to underspecify 
supertags, and show why these solutions are 
unsatisfactory. In a third part, we explain the 
solution we have adopted, building up on the 
notion of MetaGrammar introduced by Candito 
(96) and Candito (99). Finally, we discuss how 
this approach can be used in practice, and why it 
is interesting for frameworks other than LTAGs. 
1 Brief Overview of LTAGs 
A LTAG consists of a t'inite set of 
elementary trees of finite depth. Each 
elementary tree nmst "anchor" one or more 
lcxical item(s). The principal anchor is called 
"head", other anchors are called "co-heads". All 
leaves in elementary trees are either "anchor", 
"foot node" (noted *) or "substitution node" 
(noted $).. These trees are of 2 types • auxiliary 
or initial 3. A tree has at most 1 foot-node. A tree 
with a foot node is an auxiliary tree. Trees that 
are not auxiliary are initial. Elementary trees 
combine with 2 operations : substitution and 
adjunction, but we won't develop this point since 
it is orthogonal to our concern and refer to Joshi 
(87) for more details. Morphosyntactic features 
are encoded in atomic feature structures 
associated to nodes in elementary trees, in order 
to handle phenomena such as agreement. 
Moreover, linguistic constraints on the well- 
formedness of elementary trees have been 
formulated : 
• Predicate Argulnent Cooccurence Principle : 
there must be a leaf node for each realized 
argument of the head of an elementary tree. 
• Semantic consistency : No elementary tree is 
semantically void 
• Semantic minimality : an elementary tree 
corresponds at most to one semantic unit 
Figure 1 shows a non exhaustive set of 
Supertags (i.e. elementary trees) which can be 
assigned to "beats ''4 , which is a verb in trees ctl 
(canonical tree), ~2 (object extraction), 131 
(ob.iect relative) and \[32 (subject relative) and a 
noun in tree oG. So an LTAG can be seen as a 
large dictionary, were in addition of traditional 
POS, lexical entries are associated with several 
structures encoding their nlorphological as well 
as some of their syntactic properties, these 
structures being very similar to small constituent 
trees. 
e¢l c¢2 
S S 
~05 v NI,I. s' s' 
beats expl V NI,I, Cl~mp NO,I- V 
I I I I 
(Vb : "J beats M.") it is that beats 
(rb: "It is Mary that J. beats") 
N N 
NI* S' NO* S' 
N 
Comp NO$ V Camp V NI$ I I \[ I I 1,eats 
that beats who beats 
(Vb : "The man that (Vb : "The man rho (Noaa :"3 beats") 
M. beats .,.") beats 31 ...") 
HGURE 1 : some supertags fi)r "beats" 
2 Underspecifying Supertags 
The idea of underspecifying constituent trees 
(and thus elementary trees) is not new. Several 
solutions have been proposed in the past. We will 
now investigate how these solutions could 
potentially be used to encode a set of supertags in 
a compact manner. 
2.1 Parse forest 
Since elementary trees are constituent 
structures, one could represent a set of elementary 
trees with a graph instead of a tree (cf. Tomita 
(91)). This approach is not particularly interesting 
though. For example, if one considers the trees 
czl and 131 fi'om figure 1, it is obvious that they 
hardly have any structural information in 
common, not even the category of their root. 
Therefore, representing these 2 structures in a 
graph would not help. Moreover, packed 
3 Traditionally initial trees arc called a, and auxiliary lines 4 For sake of readability, morphological features arc not 
S\]IOWI1. 
447 
structures are notoriously difficult to manipulate 
and yield unreadable output. 
2.2 Logical formulae 
With this approach, developped for instance in 
Kalhneyer (99), a tree can be represented by a 
logical formula, where each pair of nodes is either 
in relation of dominance, or in relation of 
precedance. This allows to resort to 1 ~' order 
logic to represent a set of trees by 
underspecifying dominance and/or precedence 
relations . Unfortunately, this yields an output 
which is difficult to read. Also, the approach 
relies only on mathematical properties of trees 
(i.e. no linguistic motivations) 
2.3 Linear types of trees 
This approach, introduced in Srinivas (97), 
used in other work (e.g. Halber (99)) is more 
specific to TAGs. The idea is to relax constraints 
on the order of nodes in a tree as well as on 
internal nodes. A linear type consists in a 7-tuple 
<A,B,C,D,E,F,G> where A is the root of the tree, 
B is the category of the anchor, C is the lexical 
anchor, D is a set of nodes which can receive an 
adjunction, E is a set of co-anchors, F a set of 
nodes marked for substitution, and G a potential 
foot node (or nil in case the tree is initial). In 
addition, elements of E and F are marked + if 
they are to the left of the anchor, - if they are to 
the right. 
czl or2 
S S 
NO P NO,,L V PP NI,,L I /\ I /",, 
donne a N2-1- donne il N2,,\[- 
FIGURE 2 : 
two trees with the same linear type 
For example, the tree NOdonneNl'~N2 for 
"Jean donne une pomme gl Marie" (J. gives an 
apple to M.) and the tree N0donne~lN2Nl for 
"Jean donne & Marie une pomme" (J. gives M. an 
apple) which are shown on Figure 2, yield the 
unique linear type (a) 
(a) <S,V,donnc, { S,V,PP}, { h+ }, { N0-,NI +,N2+ }, nil> 
(b) <S,V,gives, { S,V,PP}, { to+ }, { N0-,N1 +,N2+} ,nil> 
This approach is robust, but not really 
linguistic : it will allow to refer to trees that are 
not initially in the grammar. For instance, the 
linear type (b) will correctly allow the sentence 
"John gives an apple to Mary", but also 
incorrectly allow "*John gives to Mary an apple". 
Moreover, linear types are not easily readable s. 
Finally, trees that have more structural 
differences than just the ordering of branches will 
yield different linear types. So, the tree 
N0giveNltoN2 (J. gives an apple to M.) yields 
the linear type (b), whereas the tree N0giveN2Nl 
(J. gives M. an apple) yields a different linear 
type (c), and thus both linear types should label 
"gives". Therefore, it is impossible to label 
"gives" with one unique linear type. 
(c) <S,V,gives, { S,V}, { }, { N0-,N 1 +,N2+} ,nil> 
2.4. Partition approach 
This approach, which we have investigated, 
consists in building equivalence classes to 
partition the grammar, each lexical item then 
anchors one class instead of a set of trees. But 
building such a partition is prohibitively costly : a 
wide coverage grammar for French contains 
approx. 5000 elementary trees (cf Abeilld & al. 
(99), (00b)), which means that we have 25~'~ 
possible subsets. Also, it does not work from a 
linguistic point of view : 
(a) Quand Jean a brisd la glace ? 
(When did J. break the ice ?) 
(b) Jean a brisd la glace (J. broke the ice) 
(c) Quelle chaise Jean a brisd ce matin ? 
(Which chair did J. break this morning ?) 
In (a) brisd potentially anchors N0briseNI 
(canonical transitive), WhN0brise (object 
extraction) and NOBriseGlace (tree for idiom). 
But in (b), we would like brim not to anchor 
WhN0brise since there is no Wh element in the 
sentence, therefore these three trees should not 
belong to the same equivallence class : We can 
have class A={N0briseN1,NOBriseGlace} and 
ClassB={WhN0brise}. But then, in (c), brisd 
potentially anchors WhN0brise and N0briseNI 
but not NOBriseGlace since glace does not appear 
in the sentence. So NOVN1 and NOBriseGlace 
should not be in the same equivalence class. This 
hints that the only realistic partition of the 
grammar would be the one were each class 
contains only one tree, which is pretty useless. 
4. Exploiting a MetaGrammar 
Candito (96), (99) has developed a tool to 
generate semi-automatically elementary trees She 
use an additional layer of linguistic description, 
called the metagrammar (MG), which imposes a 
general organization for syntactic information in 
a 3 dimensional hierarchy : 
5 This type of format was considered as a step towards 
creating a trccbank for French (of Abcilld & al 00a), but 
unfommatcly proved impossible to manually annotate. 
448 
® Dimension 1: initial subcategorization 
® Dimension 2: redistribution of functions and 
transitivity alternations 
• Dimension 3: surface realization of 
arguments, clause type and word order 
Each terminal class in dimension 1 describes a 
possible initial subcategorization (i.e. a tree 
family). Each terminal class it\] dimension 2 
describes a list of ordered redistributions of 
functions (e.g. it allows to add an argument for 
causatives). Finally, each terminal class in 
dimension 3 represents the surface realization of a 
(final) flmction (e.g. cliticized, extracted ...). 
Each class in the hierarchy corresponds to the 
partial description of a tree (cf. Rogers & Vijay- 
Shanker (94)). An elementary tree is generated by 
inheriting from one terminal class in dimension 1, 
fi'om one terminal class in dimension 2 and fl'olll 
U terulinal classes ill dinlension 3 (were n is the 
number of arguments of the elementary tree). 6 
The hierarchy is partially handwritten. Then 
crossing of linguistic phenomena (e.g. passive + 
extraction), terminal classes, and from there 
elementary trees are generated automatically off 
line. This allows to obtain a grammar which cat\] 
then be used to parse online. When the grau\]mar 
is generated, it is straight forward to keep track of 
the terminal classes each elementary tree 
inherited from : Figure 3 shows seven elementary 
trees which can superiag "domw" (gives), as well 
as the inheritance patterns 7 associated to each of 
these supertags. All the exainples below will refer 
to this figure. 
The key idea then is to represent a set of 
elementary trees by a disjunction for each 
dilnension of the hierarchy. Therefore, a hypertag 
consists in 3 disjunctions (one for dimension 1, 
one for dinlension 2 and one for dimension 3). 
The cross-product of the disiunctions can then be 
perforined automatically and from there the set of 
elementary trees referred to by the hypertag will 
6 The idea to use the MG to obtain a colnpact 
representation of a set of SuperTags was briefly sketched 
in Candito (99) and Abeill6 & al. (99), by resorting to 
MetaFeatures, but the approach here is slightly different 
since only inlbrmation about the classes in the hierarchy is 
used. 
7 We call inheritance patterns Ihe structure used to store all 
the terminal classes a tree has inherited from. 
{ ) CA ..... ...... 
oq 
,~ V~)i ....... ioII 1 : It0 '¢li i ('Ill 2) q 
N~'P ~ II,i ....... i,,, 2 ....... lislril,,,,ioi, .1 
\[~)imension3: suhj:tlominai.canonlcal \[\[ 
\[ //~.x ollj ....... ina, ........ |cal II 
donne h N2-L ~l'O )j : II(lltl IIn -canon efll\]\[ 
(J donne lille pOlllMt~ h M. / 
J gives an apple to M) 
c~2 
N~N I.L \[~)i ....... ioi, 1: n0vnl (hi12) q I Dimc.sio. 2 : .o redistril>ution t 
I lii ....... ion 3 :\] suhj ....... inal ........ ieal II ~ / I obJ ....... i,,,,I ........ i~.l II 
(J, donlIc \[I ~.~. \[\]Dt3 pOlllmO / 
J gi','es to M an apll\]¢) 
0'.~ F) i ....... ion |: I, OVlll(~ln2) "~1 
s ~ I Dimension 2 : hObj-clnply / 
N~NI.L~ \] I)i ....... ion 3 :\[ sulij ....... inal ........ ical I\[ I 
L_ I ebj : nominal-canonical I\] 
define 
(J. dOlitle LilLe pOlliille/ 
J gives an apple) 
\[;4 ~li ....... ioi, I : nll~ nl(hi,2) 
~~ I I)hllensi(lll 2 : no redistribution \[ 
Conll~ %' Nll$ PP \[ I obj : relallvizcd-object II I I /\ L i.o,,J: 
...... ,,,,,, ........ ++.~ 
que donne l'\[ep N2.\[- 
/ a 
(1 peru n • que donne J. h M, I 
The allph, wDich gives J. to M.) 
t\]5 \[~)i ....... ioi, 1: n0viil(hn2) q N 
NI*~S_' ~ I Dime.d.n 2 :no redistril}utlon .1 
~~ l l)i ....... i0113 :l stlhj ........ ilHll ......... i'l II 
Cotnl> N0-L V PP l \] obj : rehttivized-olljcct II 
I I ~ L l,,-.,,J ........ i,,,,: .......... i,.,~ que donne P\[cp N2,~ 
/ /l 
(l,a pomme que J. donne D M. I 
The allph, whictl J. gil'es to M.J 
I}6N ~)i ....... ion I: n0vnl(',,,2) q 
I m ....... io,, 2 :aohj-cmply t NI* 
-- ~ ~ II)i ....... ion 3 :l suhj ........ iual-i.verled \]l / A 
Comp V NI)~ L i ob.i : relali;ized-object .~J I i 
qlte dotlttd 
(La pomme que donne J. / 
The apph' which give.i J.) 
\[:17 UIi ....... ion 11 n(l',nl(hn2) q 
y s' ~ I I)imc.sio. 2 : Mll, j-elnpty .l NI* l m ....... io. 3 :l +,,i,j ........ inal 
.......... it,,: \]I c,,~ v L 
I obj : relali+ized.ohject 11 
I i 
qtte donne 
(La potlltJl# qtte J. dotltte / 
The allPh' which J. gil.es ) 
FIGURE 3 : SuperTags and associated 
inheritance patterns 
be automatically retrieved We will now ilhlstrate 
this, first by showing how hypertags are built, and 
then by explaining how a set of trees (and thus of 
supertags) is retrieved from the information 
contained in a hypemig. 
4.1 Building hypertags : a detailed example 
Let us start with a simple exemple were we 
want "donner" to be assigned the supertags o~1 (J. 
dmme tree pomme D M.I J. gives an apple to M.) 
and o~2 (J donne h M. tree l)omme/J, gives M. an 
449 
apple). On figure 3, one notices that these 2 trees 
inherited exactly fi'om the same classes : the 
relative order of the two complements is left 
unspecified in the hierarchy, thus one same 
description will yield both trees. In this case, the 
hypertag will thus simply be identical to the 
inheritance pattern of these 2 trees : 
Dimension 1 : n0vnl (hn2) 
Dimension 2 : no redistribution 
Dimension 3 subj :nominal-canonical \[ 
obj : nominal-canonical \] 
\[ a-~'}bj: nominal-canonical\[ 
Let's now add tree o{3 (J. donne une pomme / 
J. gives an apple) to this hypertag. This tree had 
its second object declared empty in dimension 2 
(thus it inherits only two terminal classes from 
dimension 3, since it has only 2 arguments 
realized). The hypertag now becomes 8 : 
Dim. 1: n0vnl(an2) 
Dim. 2 : no redistribution OR StObj- empty 
I)im. 3 lsubj :nonainal-canonical \[ 
obj : nominal-canonical 
a-obj: nominal-canonical 
Let's now add the tree 134 for the object 
relative to this hypertag. This tree has been 
generated by inheriting in dimension 3 fi'om the 
terminal class "nominal inverted" for its subject 
and from the class "relativized object" for its 
object. This information is simply added in the 
hypertag, which now becomes : 
I )i,l~. : n0wll (~.12) ira. 2 : no redistribution 0P, il0bj- empty l ira. 3 subj :nominal-canonical OR nominal-inverledl I obj : nominal-canonical OR relativized-oblect I I I a-0bj: n0minal-canonical ii 
Also note that for this last example the 
structural properties of 134 were quite different 
than those of ¢~1, 0{2 and cG (for instance, it has a 
root of category N and not S). But this has little 
importance since a generalization is made in 
linguistic terms without explicitly relying on the 
shape of trees. 
it is also clear that hypertags are built in a 
monotonic fashion : each supertag added to a 
hypertag just adds information. Hypertags allow 
to label each word with a unique structure 9. and 
8 What has been added to a supertag is shown in bold 
characters. 
9 We presented a simple example for sake of clarity, but 
traditional POS ambiguity is handled in the same way, 
except that disjunctions are then added in dimension 1 as 
contain rich syntactic and ftmctional information 
about lexical items (For our example here the 
word donne~gives). They are linguistically 
motivated, but also yield a readable output. They 
can be enriched or modified by Imman annotators 
or easily fed to a parser or shallow parser. 
4.2 Retrieving information from hypertags 
Retrieving inforlnation from hypertags is 
pretty straightforward. For example, to recover 
the set of supertags contained in a hypertag, one 
just needs to perform the cross-product between 
tile 3 dimensions of the hypertag, as shown orl 
Figure 4, in order to obtain all inheritance 
patterns. These inheritance patterns are then 
matched with tile inheritance patterns contained 
in the grammar (i.e. tile right colunm in Figure 3) 
to recover all the appropriate supertags. 
Inheritance patterns which are generated but don't 
match any existing trees in tile grammar are 
simply discarded. 
We observe that the 4 supertags 0{1, c~2 and 
0{3 and \]34 which we had explicitly added to tile 
hypertag in 4.1 are correctly retrieved. But also, 
the supertags 135, 136 and 137 arc retrieved, which 
we did not explicitly intend since we never added 
them to the hypertag. But if a word can anchor 
the 4 first trees, then it will also necessarily 
anchor tile three last ones : for instance we had 
added the canonical tree without a second object 
realized into the hypertag (tree or2 ), as well as 
the tree for tile object relative with a second 
object realized realized (tree 134 ), so it is 
expected that tile tree for the object relative 
without a second object realized can be retrieved 
from the hypertag (tree 136) even though we never 
explicitly added it. In fact, the automatic crossing 
of disjunctions in the hypertag insures 
consistency. 
Also note that no particular" mechanism is 
needed for dimension 3 to handle arguments 
which are not realized : if hObj-empty is inherited 
from dilnension 2, then only subject and object 
will inherit from dimeusiou three (since only 
arguments that are realized inherit from that 
dimension when the grammar is generated). 
Information can be modified at runtime in a 
hypertag, depending on the context of lexical 
items. For example "relativized-object" can be 
suppressed in dimension 2 from the hypertag 
shown on Figure 4, in case no Wh element is 
encountered in a sentence. Then, the correct set 
of supertags will still be retrieved from the 
well. 
450 
Content of the llypertag 
Dim ension2 Dim en sio n3 
Subject Object a-obj 
1 iI 
l .-/\ ~ ~ 1 1 1 i 1 ~1 0<2 \[~5 N o \[~ cs3 \[~6 N. \[7,7 
(Jorresllonding Corresl)onl(lhig 
I,'~'~' \[%qsertagsc°r-resls°'idi"glol'il-ierilancelJatter'sl t r~,'e 
\[ (el Figure 3) I 
FIGURE 4 : Retrieving Inheritance patterns and Supertags 
fronl a Hypertag 
hypertag by automatic crossing (that is, trees o~1, 
(;~2 and o'.3), since the other inheritance l)atterns 
generated won't refer to any tree ill the grainmar 
(here, tie tree inherits in diillension 3 
,vuhject:in, verted-nominal, without inheriting also 
objecl: IwlalivizeU-oluect) 
4.3 Practical use 
We have seen that an LTAG can be seen as a 
dictionary, in which each lexical entry is 
associated to a set of elementary trees. With 
hypertags, each lexical entry is now paired with 
one unique structure. Therefore, automatically 
hypertagging a text is easy and involves a simple 
dictionary lookup. The equiwllent of finding the 
"right" supertag for each lexical item in a lext (i.e. 
reducing ambiguity) then consists in dynamically 
removing information from hypertags (i.e. 
suppressing elements in disjunctions). This can be 
achieved by specific rules, which are currently 
being developed. The resulting output carl then 
easily be manually annotated in order to build a 
gold-standard corpus : manually removing 
linguistically relevant pieces fronl information in 
a disjunction from a single structure is simpler 
than dealing with a set of trees. In addition of 
obvious advantages in terms of display (tlee 
structures, especially when presented in a non 
graphical way, are unreadable), the task itself 
becomes easier because topological problems are 
solved automatically: annotators need just 
answer questions such as "does this verb have an 
extracted object ?", "is the subject of this verb 
inverted ?" to decide which terminal classe(s) 
nlust be kept i° .We believe that these questions 
are easier to iulswcr than "Which of these trees 
have a node N I marked wh+ at address 1.1 9" 
(for an extracted object). 
Moreover, supertagged text are difficult to use 
outside of an LTAG framework, contrary to 
hypertagged texts, which contain higher level 
general linguistic information. An example would 
be searching and extracting syntactic data oil a 
large scale : suppose one wants to extract all tile 
occurrences where a given verb V has a 
relativized object. To do so on a hypertagged text 
simply involves performing a "grep" ell all lines 
coutainhig a V whose hypertag contains 
dimension .7 : objet:relalivized-object , without 
knowing anything about the LTAG framework. 
Performing the same task with a supertagged text 
involves knowing how LTAGs encode relativized 
objects in elementary trees and scanning potential 
trees associated with V. Another examl)le would 
be using a hypertagged text as an input to a parser 
based oil a framework other than LTAGs : for 
instance, information in hypertags could be used 
by an LFG parser to constrain the construction of 
an IV-structure, whereas it's uuclear how tills could 
be achieved with supertags. 
10 This of course implies that one must be very careful in 
choosing evocative names for terminal classes. 
451 
The need to "featurize" Supertags, in order to 
pack ambiguity and add functional information 
has also been discussed for text generation ill 
Danlos (98) and more recently in Srinivas & 
Rainbow (00). It would be interesting to compare 
their approach with that of hypertags. 
Conclusion 
We have introduced the notion of Hypertags. 
Hypertags allow to assign one unique structure to 
lexical items. Moreover this structure is readable, 
linguistically and computationally motivated, 
and contains much richer syntactic information 
than traditional POS, thus a hypertagger would be 
a good candidate as the front end of a parser. It 
allows in practice to build large annotated 
resources which are useful for extracting 
syntactic information on a large scale, without 
being dependant on a ~iven grammatical 
formalism. 
We have shown how hypertags are built, how 
information can be retrieved from them. Further 
work will investigate how hypertags can be 
combined directly. 

Referenees 

Abeilld A., Candito M.H., Kinyon A. (1999) FFAG : 
current status & parsing scheme. Prec. Vextal'99. 
Venice. 

Abeilld A., Cldment L., Kinyon A. (2000a) Building a 
Treebank for French. Prec. LREC'2000. Athens. 

Abeilld A., Candito M.H., Kinyon A. (2000b) Current 
status of FTAG. Proc TAG+5. Paris. 

Barrier N. Barrier S. Kinyon A. (2000). Lcxik : a 
maintenance tool for FTAG. Prec. TAG+5. Paris. 

Candito M-H. (1996) A principle-based hierarchical 
representation of LTAGs, Prec. COLING'96 
Kopenhagen. 

Candito M.-H, (1999) Reprdsentation modulairc ct 
paramdtrable de grammaircs 61ectroniques 
lexicalisdes. Application au frangais eta l'italien. 
PhD dissertation. University Paris 7. 

Chen J., Srinivas B., Vijay-Shanker K. 1999 New 
Models for Improving Supertag Disambiguation. 
Prec. EACL'99 pp. 188-195. Bergen. 

Cldment L, Kinyon A. (2000) Chunking, marking and 
searching a morphosyntactically annotated corpus 
for French. Prec. ACIDCA'2000. Monastir. 

Danlos L (1998) GTAG : un formalisme lexicalis6 
pour la gdndration automatique de TAG. TAL 39:2. 

Giguet E. (1998) Mdthodes pour l'analyse automatique 
de structures formelles sur documents 
multilingues.PhD thesis. Universitd de Cacn. 

Halber A. (1999) Stratdgie d'analyse pour la 
comprdhension de la parole : vers une approche 
base de Grammaires d'Arbres Adjoints Lexicalisdes. 
PhD thesis. ENST. Paris 

Joshi A. (1987) An introduction to Tree Adjoining 
Grammars. In Mathematics of Language. A. 
Manaster-Ramer (eds). John Benjamins Publishing 
Company. Amsterdam.Philadelphia. pp. 87-114. 

Joshi A. (1999) Explorations era domain of locality. 
CLIN'99. Utrecht. 

Joshi A. Srinivas B. (1994) Disambiguation of Super 
Parts of Speech (or Supertags) : Ahnost parsing. 
Proceeding COLING'94. Kyoto. 

Kalhneyer L 0999) Tree Description Grammars and 
Undcrspecified Representations. PhD thesis, 
Universit'At Ti.ibingen. 

Kinyon A. (1999a) Parsing preferences with LTAGs : 
exploiting the derivation tree. Prec. ACL'99.College 
Park, Md 

Kinyon A. (1999b) Some remarks about the 
psycholinguistic relevance of LTAGs. Prec. 
CLIN'99. Utrecht. 

Srinivas B. (1997) Complexity of lexical descriptions 
and its relevance tbr partial parsing, PhD thesis, 
Univ. of Pennsylvania. 

Srinivas B., Joshi A. (1999) Supertagging : An 
approach to almost parsing. Computational 
Linguistics 25:2. 

Srinivas B. Rainbow O. (2000) Using TAGs, a Trce 
Model, and a Language Model for Generation. Prec. 
TAG+5. Paris. 

Tomita M. (1991) Generalized LR Parsing. Masaru 
Tomita (eds). Kluwer academic publishers.. 

Rogers J., Vijay-Shanker K. (1994) Obtaining trees 
froln their descriptions : an application to TAGs. 
Computational Intelligence, 10:4 pp 401-421. 

Voutilainen A. Tapanainen P. (1993) Ambiguity 
resolution in a reductionistic parser. Prec. EACL'93. 
