Where we are
Yesterday:
Working with Tree-Adjoining Grammar
introducing XTAG analyses of extraction
Day 4: Grammar implementation & metagrammars
di�erent templates for di�erent constructions and valency
frames
rich feature sets in the nodes (for constraining substitution and
adjunction)
Kata Balogh & Timm Lichte
University of Düsseldorf, Germany
Today:
How to factorize the set of templates?
ESSLLI 2015, Barcelona
) express lexical generalizations, e.g. active-passive diathesis
) define tree families
SFB 991
How to turn this into an electronic resource?
How to plug it into a lexicon and use it?
Balogh & Lichte (Düsseldorf)
Outline
1
What is grammar implementation?
2
Two ways of tree template implementation
Metarules
Metagrammars
3
eXtensible Metagrammar (XMG)
4
LTAG as Construction Grammar
5
Lexicon and parser
6
Two kinds of grammar implementation
grammar/
linguistic theory
“implementation”
specifications
in accordance with a
grammar formalism
evaluation
of the theory
As is frequently pointed out but cannot be overemphasized, an important goal
of formalization in linguistics is to enable subsequent researchers to see the
defects of an analysis as clearly as its merits; only then can progress be made
e�iciently. (Dowty 1979: 322)
Summary
Balogh & Lichte (Düsseldorf)
2
3
Balogh & Lichte (Düsseldorf)
4
Two kinds of grammar implementation
What kind of grammar resource?
grammar/
linguistic theory
S
“implementation”
tree template
specifications
in accordance with a
grammar formalism
NP
VP
V⇧
NP
“implementation”
evaluation
of the theory
grammar resource
lexical insertion
anchor
computational
application
Balogh & Lichte (Düsseldorf)
4
The implementation task for LTAG
repairs
Balogh & Lichte (Düsseldorf)
5
Two ways of grammar implementation with TAG
Two existing toolkits:
XTAG tools[27]
1 implementation tools
General task
Implement a large-coverage LTAG, i.e. based on the XTAG grammar!
) metarule approach
Subtasks:
1
Generate unlexicalized trees (= tree templates)!
2
Generate a database of lexical anchors (= the lexicon)!
3
Connect the tree templates with the lexicon (= lexical insertion)!
Balogh & Lichte (Düsseldorf)
2
editor/viewer for MorphDB and SynDB
3
parser
XMG + lexConverter + TuLiPA
1
XMG: eXtensible MetaGrammar[9]
) metagrammar approach
6
2
lexConverter (LEX2ALL)
3
TuLiPA: Tübingen Linguistic Parsing Architecture[18]
Balogh & Lichte (Düsseldorf)
7
Outline
The situation
12 templates
for intransitive verbs
1
What is grammar implementation?
2
Two ways of tree template implementation
Metarules
Metagrammars
3
eXtensible Metagrammar (XMG)
4
LTAG as Construction Grammar
5
Lexicon and parser
39 tree templates
for transitive verbs
S
S
NP
VP
NP
VP
V
V
S
NP
S
NP
S
NP
NP
S
VP
NP
VP
V
6
V
...
Summary
NP
...
Basically, XTAG defines a set of 1008 unrelated tree templates.
Balogh & Lichte (Düsseldorf)
8
Metarules for LTAG
Balogh & Lichte (Düsseldorf)
9
Metarules for LTAG: Example
S
S
Idea from GPSG[12] , later applied to XTAG[2,3,21]
extraction:
=)
NP
NP
S
NP
tree fragments
metarules
accumulation
S
core grammar
(tree templates)
metarules
expanded grammar
(tree templates)
NP1
S
passivization:
NP0
VP
VP
=)
VP
metarules connect tree templates of a
tree family
VP
V⇧
NP1
V⇧
PP
P
NP0
by
Balogh & Lichte (Düsseldorf)
10
Balogh & Lichte (Düsseldorf)
11
Metarules for LTAG: Problems[2]
Metarules for LTAG: Example
Metarules are very powerful:
nx0Vnx1
extraction
deletion, copying, recursive application, metavariables over trees
passivization
order sensitive
in the unrestricted case: undecidable[25]
nx1Vbynx0
W0nx0Vnx1
extraction
Restrictions (GPSG):[22]
finite closure: apply every metarule at most once!
W1nx1Vbynx0
) still NP-complete
Tnx0nx1
biclosure: apply at most two metarules in a row!
) insu�icient for LTAG metarules[2]
explicit rule ordering (by means of finite state automata)[21]
Balogh & Lichte (Düsseldorf)
12
Metagrammars for LTAG
13
Metagrammars for LTAG: Tree descriptions
LD : Description language for trees
Candito (1996)[8,9,26]
Let n2 and n2 be node variables:
accumulation of
descriptions
tree fragments
Balogh & Lichte (Düsseldorf)
n ! n2 | n1 !+ n2 | n1 !⇤ n2 |
*. 1
+
n
n2 |
n1 + n2
| n1 ⇤ n2 | //
Description := .. 1
//
. n1 = n2 |
, Description ^ Description
tree templates
Example:
S
arbitrary
disjunction
NP
corresponds to
VP
S
tree families
NP
Balogh & Lichte (Düsseldorf)
14
Balogh & Lichte (Düsseldorf)
⇤
+
NP
corresponds to
nS ! nNP
nS ! nVP
nNP nVP
^
^
nS ! nNP1
^
nS !+ nNP2 ^
nNP1 ⇤ nNP2
15
Metagrammars for LTAG: Example
Metagrammars for LTAG: Example
Minimal model of tree descriptions
Minimal model of tree descriptions
You may add edges but not nodes!
You may add edges but not nodes!
S
S
NP
NP
S
S
S
S
NP
NP
|=
VP
VP
S
NP
VP
V⇧
VP
VP
V⇧
NP
NP
Balogh & Lichte (Düsseldorf)
16
Metagrammars for LTAG: Example
16
You may add edges but not nodes!
S
Balogh & Lichte (Düsseldorf)
|=
NP
NP
S
VP
VP
V⇧
NP
NP
S
VP
NP
VP
S
NP
S
NP
VP
V⇧
VP
S
S
NP
NP
S
VP
NP
VP
NP
|=
S
VP
NP
Minimal model of tree descriptions
S
NP
V⇧
Balogh & Lichte (Düsseldorf)
You may add edges but not nodes!
S
VP
Metagrammars for LTAG: Example
Minimal model of tree descriptions
NP
VP
|=
NP
Balogh & Lichte (Düsseldorf)
...
S
NP
16
VP
VP
17
Metagrammars for LTAG: Properties
Metagrammars for LTAG: Passivization
S
no deletion, no copying, no recursion
NP
declarative, order insensitive
The number of minimal models is finite.
^
VP
VP
^
V⇧
disjunction
does the trick!
BUT: the number of minimal models can grow exponentially
(O(n!)) in terms of the number of described nodes.
VP
*.
..
..
VP
..
..
.. V⇧ NP
..
.
,
V⇧
_
P
S
NP
|=
NP
VP
V⇧
VP
V⇧
NP
by
S
Does it su�ice? How to express passivization?
PP
+/
//
//
//
//
//
//
/
-
PP
P
NP
NP
by
Tnx0nx1
Balogh & Lichte (Düsseldorf)
18
Metagrammar for LTAG: Classes
Balogh & Lichte (Düsseldorf)
19
Metagrammar for LTAG: Classes
Tree descriptions are bundled into so-called classes:
LC : Description language for the combination of tree descriptions
Tnx0Vnx1:
Class := Name : Content
Description | Name |
*
Content := .. Content _ Content |
, Content ^ Content
Upon instantiating/using a class:
NP
+/
/
-
VP
Tnx0Vnx1: Subject
Node variables are replaced by fresh ones.
Tnx0Vnx1:
Node variables are known to the instantiating class.
^
^
VP
V⇧
^
VP
V⇧
_
+/
//
//
PP
//
//
P
NP //
//
/
by
-
VerbProjection ^ (Object _ by-Phrase)
nx0V
Subject
The class name is replaced by the content in the instantiating
class.
VerbProjection
^ (Object _ by-Phrase)
Object
by-Phrase
nx0V
) Classes can be reused!
Balogh & Lichte (Düsseldorf)
S
*.
..
..
VP
..
..
.. V⇧ NP
..
.
,
Tnx0Vnx1
20
Balogh & Lichte (Düsseldorf)
21
Metagrammar for LTAG: Class hierarchies
Metagrammar for LTAG: Class hierarchies
There are very many possible class hierarchies . . .
Subject
There are very many possible class hierarchies . . .
WhNP+EmptyWord
BaseSubject
Subject
Object
WhSubject
WhObject
BaseObject
Object
nx0V
VerbProjection
alphanx0V
VerbProjection
alphanx0V
WhNP+EmptyWord
alphaW0nx0V
nx0Vnx1
Wnx0Vnx1 alphanx0Vnx1
alphaW0nx0V
alphaW0nx0Vnx1
Tnx0V
alphaW0nx0Vnx1
alphaW1nx0Vnx1
alphanx0Vnx1
alphaW1nx0Vnx1
Tnx0V
Tnx0Vnx1
Tnx0Vnx1
Balogh & Lichte (Düsseldorf)
22
Metagrammar for LTAG: Class hierarchies
There are very many possible class hierarchies . . .
WhNP+EmptyWord
WhSubject
Subject
Subject
WhObject
VerbProjection
22
Metagrammar for LTAG: Class hierarchies
There are very many possible class hierarchies . . .
BaseSubject
Balogh & Lichte (Düsseldorf)
BaseSubject
BaseObject
WhNP+EmptyWord
WhSubject
Object
WhObject
BaseObject
VerbProjection
Object
Tnx0V
Tnx0V
Balogh & Lichte (Düsseldorf)
Tnx0Vnx1
Tnx0Vnx1
[1]
22
Balogh & Lichte (Düsseldorf)
22
Metagrammar for LTAG: Class hierarchies
Outline
. . . but not everything is possible:
alphanx0Vnx1
alphaW0nx0Vnx1
1
What is grammar implementation?
2
Two ways of tree template implementation
Metarules
Metagrammars
3
eXtensible Metagrammar (XMG)
4
LTAG as Construction Grammar
5
Lexicon and parser
6
Summary
alphanx1Vbynx0
alphaW1nx1Vbynx0
Tnx0nx1
Balogh & Lichte (Düsseldorf)
23
eXtensible Metagrammar (XMG): Background
Balogh & Lichte (Düsseldorf)
eXtensible Metagrammar (XMG): Example
S
developed at LORIA, Nancy, and LIFO, Orléans.[9]
wri�en in Oz/Mozart YAP and Python (as of XMG2)
NP#
available at http://sourcesup.cru.fr/xmg
Why “eXtensible” ?
highly
VP
class Subject
export ?S
modularized[19]
declare ?S ?NP ?VP
dimensions with dedicated description languages and compilers
(<syn>, <sem>, <frame>, <morph>, . . . )
{ <syn>{
node ?S [cat=s]{
interface through shared variables
node ?NP (mark=subst) [cat=np]
node ?VP [cat=vp]
Some existing implementations using XMG:
}
French: FrenchTAG[8]
}
English: XTAG with XMG[1]
German:
Balogh & Lichte (Düsseldorf)
24
}
GerTT[15]
25
Balogh & Lichte (Düsseldorf)
26
eXtensible Metagrammar (XMG): Example
eXtensible Metagrammar (XMG): Example
S
S
NP#
VP
VP
NP#
VP
V⇧
|=
S
NP#
class Subject
VP
V⇧
export ?S
declare ?S ?NP ?VP
class alphanx0v
{ <syn>{
import VerbProjection[]
node ?S [cat=s];
declare ?Subj
node ?NP (mark=subst) [cat=np];
{
node ?VP [cat=vp];
?Subj = Subject[];
?S -> ?NP;
?Subj.?VP = ?VP
?S -> ?VP;
}
?NP >> ?VP
}
}
Balogh & Lichte (Düsseldorf)
27
eXtensible Metagrammar (XMG): Example with <frame>
(Lichte &
Outline
...
<syn>{
node ?S [cat=s];
node ?SUBJ [cat=np,
top=[i=?1]];
S
NP[I= 1 ]
VP[E= 0 ]
V⇧[E= 0 ]
26
�����
0 66
64�����
Balogh & Lichte (Düsseldorf)
37
77
17
5
28
Petitjean)[17]
class Subj
class Subj
Balogh & Lichte (Düsseldorf)
node ?VP [cat=vp,bot=[e=?0]];
node ?V (mark=anchor)
[cat=v,top=[e=?0]];
?S -> ?SUBJ; ?S -> ?VP; ?VP -> * ?V;
?SUBJ >> ?VP
};
<frame>{
?0[event,
actor:?1]
}
...
29
1
What is grammar implementation?
2
Two ways of tree template implementation
Metarules
Metagrammars
3
eXtensible Metagrammar (XMG)
4
LTAG as Construction Grammar
5
Lexicon and parser
6
Summary
Balogh & Lichte (Düsseldorf)
30
LTAG as Construction Grammar
LTAG as Construction Grammar
By virtue of adjunction, cases of long-distance dependencies can be
immediately captured:
Claim
(ongoing work)
(1) Who does Mary say sometimes walks into the house.
LTAG shares central ideas with (some versions of) Construction Grammar
(CxG):[13]
S
1
only surface structure: no transformational or derivational component
AUX
2
grammatical constructions: “form-function pairings at varying
levels of complexity and abstraction” (e.g. words, phrases)
does
3
a network of constructions “which nodes are related by inheritance links ”
S
S
NP
NP
S
VP
V
NP
S*
VP
VP
say
PP
V
walks
only surface structure
Balogh & Lichte (Düsseldorf)
31
LTAG as Construction Grammar
X
Balogh & Lichte (Düsseldorf)
32
LTAG as Construction Grammar
S
NP[�=
1
0
]
]
VP[�=
0
]
0
VP[�=
V⇧[�=
0
]
0
]
NP[�= 2 ]
PP[����=to, �= 3 , �= 4 ]
VP
VP
(2) John gives/sends the book to Mary
S[�=
VP
NP
VP
V⇧
26���������
37
66
77
2
3
66
77
66activity 77
66����� 6
77
7
17
�����
6
66
7
4
5
26
3777
66
bounded-translocation
66
7777
66
7777
66������ 4 66����� 2
66����
7777
66
3
4
575
4
V⇧
⇤
Subj
V⇧
VSpine
DirObj
DirPrepObj
intransitive
intransitive
motion construction
X
transitive
transitive
motion construction
inheritance network of constructions
Balogh & Lichte (Düsseldorf)
PP
NP
V⇧
(from Kallmeyer & Osswald (2013))
grammatical constructions
VP
33
Balogh & Lichte (Düsseldorf)
PO-to
X
34
LTAG as Construction Grammar: Summary
LTAG incorporates central ideas of CxG:
only surface structure
X
X
Outline
1
What is grammar implementation?
2
Two ways of tree template implementation
Metarules
Metagrammars
3
eXtensible Metagrammar (XMG)
4
LTAG as Construction Grammar
5
Lexicon and parser
6
Summary
grammatical constructions
inheritance network of constructions
X
LTAG di�ers substantially from other implementations of CxG
(e.g. Sign-based CxG[23] , Fluid CxG[24] ).
) di�erent empirical predictions or theoretical ramifications?
Balogh & Lichte (Düsseldorf)
35
Lexicon and parser
Balogh & Lichte (Düsseldorf)
36
Lexicon and parser: A 2-layered lexicon
loves
love
“morphological lexicon”
metagrammar
XMG compiler
Tnx0Vnx1
“lemma lexicon”
implementational
cycle
Morphological lexicon
maps an (inflected) token to some base form (= lemma), while preserving
morphological information in a feature structure.
compiled
metagrammar
2-layered lexicon
TuLiPA parser
loves
Peter
[pos=v; num=sing; pers=3;]
[pos=n; num=sing; pers=3; case=nom|acc;]
parsing result
Interface with tree templates: Feature unification during lexical
insertion
input sentence
Balogh & Lichte (Düsseldorf)
love
Peter
37
Balogh & Lichte (Düsseldorf)
38
Lexicon and parser: A 2-layered lexicon
Lexicon and parser: The TuLiPA parser
(Parmentier et al.)[18]
loves
love
“morphological lexicon”
Tnx0Vnx1
TuLiPA
Tübingen Linguistic Parsing Architecture (TuLiPA)
“lemma lexicon”
uses Range Concatenation Grammar (RCG) as a pivot formalism.
Lemma lexicon
maps a lemma onto tree tuple families, while also containing selectional
restrictions (e.g., case assignment).
*ENTRY: love
*CAT: v
*SEM:
*ACC: 1
*FAM: Tnx0Vnx1
*FILTERS: []
*EX:
*EQUATIONS:
NParg1 -> case = nom
NParg2 -> case = acc
*COANCHORS:
Components:
What is grammar implementation?
2
Two ways of tree template implementation
Metarules
Metagrammars
3
eXtensible Metagrammar (XMG)
4
LTAG as Construction Grammar
5
Lexicon and parser
6
Summary
2
RCG parser ! RCG derivation forest ! TAG derivation forest
Parse viewer (derived tree, derivation tree, dependency view,
semantic representation)
Availability of TuLiPA:
wri�en in Java and released under the GNU GPL
(http://sourcesup.cru.fr/tulipa/)
38
Outline
1
TAG-to-RCG converter (on-line)
3
Interface with tree templates:
EQUATIONS ! nodes of tree templates
FILTERS ! selection of tree templates
Balogh & Lichte (Düsseldorf)
1
Balogh & Lichte (Düsseldorf)
39
Summary
A metagrammar contains descriptions of unanchored elementary
trees.
Metagrammar descriptions are declarative and multidimensional.
Metagrammar descriptions make up an inheritance hierarchy.
The metagrammar allows one to express and implement lexical
generalizations, e.g. active-passive diathesis.
TAG + metagrammars implements central ideas of CxG.
Hot topics:
parsing with metagrammars[7]
use metagrammars for morphological descriptions[11,20]
Adjacent topics:
grammar induction from treebanks[5,6,14,26]
Balogh & Lichte (Düsseldorf)
40
Balogh & Lichte (Düsseldorf)
41
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
Alahverdzhieva, Katya. 2008. XTAG using XMG. A core Tree-Adjoining Grammar for
English. University of Nancy 2 / University of Saarland Master’s Thesis.
http://homepages.inf.ed.ac.uk/s0896251/pubs/msc-sb2008.pdf.
Becker, Tilman. 1994. Hytag: a new type of tree adjoining grammars for hybrid syntactic
representations of free word order languages. Universität des Saarlandes PhD thesis.
http://www.dfki.de/~becker/becker.diss.ps.gz.
Becker, Tilman. 2000. Pa�erns in metarules for TAG. In Anne Abeillé & Owen Rambow
(eds.), Tree Adjoining Grammars: Formalisms, linguistic analyses and processing (CSLI
Lecture Notes 107), 331–342. Stanford, CA: CSLI Publications.
Candito, Marie-Hélène. 1996. A principle-based hierarchical representation of LTAGs.
In Proceedings of the 16th international Conference on Computational Linguistics
(COLING 96). Copenhagen. http://aclweb.org/anthology-new/C/C96/C96-1034.pdf.
Chen, John, Srinivas Bangalore & K. Vijay-Shanker. 2006. Automated extraction of
Tree-Adjoining Grammars from treebanks. Natural Language Engineering 12 (3).
251–299.
Chiang, David. 2000. Statistical parsing with an automatically-extracted Tree Adjoining
Grammar. In Proceedings of the 38th annual meeting of the Association for
Computational Linguistics, 456–463. Hong Kong.
de la Clergerie, Éric Villemonte. 2013. Exploring beam-based shi�-reduce dependency
parsing with DyALog: results from the SPMRL 2013 shared task. In 4th workshop on
statistical parsing of morphologically rich languages (SPMRL’2013). Sea�le.
http://hal.inria.fr/docs/00/87/91/29/PDF/dyalogsr.pdf.
Crabbé, Benoît. 2005. Représentation informatique de grammaires d’arbres fortement
lexicalisées: Le cas de la grammaire d’arbres adjoints. Université Nancy 2 PhD thesis.
Kallmeyer, Laura, Timm Lichte, Wolfgang Maier, Yannick Parmentier &
Johannes Dellert. 2008. Developing a TT-MCTAG for German with an RCG-based
parser. In European Language Resources Association (ELRA) (ed.), Proceedings of the
sixth international Conference on Language Resources and Evaluation (LREC’08).
Marrakech, Morocco.
Kallmeyer, Laura & Rainer Osswald. 2013. Syntax-driven semantic frame composition
in Lexicalized Tree Adjoining Grammar. Journal of Language Modelling 1. 267–330.
Lichte, Timm & Simon Petitjean. 2015. Implementing semantic frames as typed feature
structures with XMG. Journal of Language Modelling 3(1). 185–228.
http://jlm.ipipan.waw.pl/index.php/JLM/article/view/96.
Parmentier, Yannick, Laura Kallmeyer, Wolfgang Maier, Timm Lichte &
Johannes Dellert. 2008. TuLiPA: A syntax-semantics parsing environment for mildly
context-sensitive formalisms. In Proceedings of the ninth international workshop on Tree
Adjoining Grammars and related formalisms (TAG+9), 121–128. Tübingen, Germany.
Petitjean, Simon. 2014. Génération Modulaire de Grammaires Formelles. Orléans,
France: Université d’Orléans Thèse de Doctorat.
https://tel.archives-ouvertes.fr/tel-01163150/.
Petitjean, Simon, Younes Samih & Timm Lichte. 2015. Une métagrammaire de
l’interface morpho-sémantique dans les verbes en arabe. In Actes de la 22e conférence
sur le Traitement Automatique des Langues Naturelles, 473–479. Caen, France.
http://www.atala.org/taln_archives/TALN/TALN-2015/taln-2015-court-024.
Prolo, Carlos A. 2002. Generating the XTAG English grammar using metarules. In
Proceedings of the 19th international Conference on Computational Linguistics (COLING
2002), 814–820. Taipei. Taiwan.
[9]
[10]
[11]
[12]
[13]
[14]
[22]
[23]
[24]
[25]
[26]
[27]
Crabbé, Benoit, Denys Duchier, Claire Gardent, Joseph Le Roux & Yannick Parmentier.
2013. XMG: eXtensible MetaGrammar. Computational Linguistics 39(3). 1–66.
http://hal.archives-ouvertes.fr/hal-00768224/en/.
Dowty, David R. 1979. Word meaning and Montague Grammar. Reprinted 1991 by
Kluwer Academic Publishers. Dordrecht: D. Reidel Publishing Company.
Duchier, Denys, Brunelle Magnana Ekoukou, Yannick Parmentier, Simon Petitjean &
Emmanuel Schang. 2012. Describing morphologically rich languages using
metagrammars: A look at verbs in Ikota. In Workshop on language technology for
normalisation of less-resourced languages (SALTMIL 8 – AfLaT 2012), 55–59.
http://www.tshwanedje.com/publications/SaLTMiL8-AfLaT2012.pdf#page=67.
Gazdar, Gerald. 1981. Unbounded dependencies and coordinated structure. Linguistic
Inquiry 12. 155–182.
Goldberg, Adele. 2013. Constructionist approaches. In Thomas Ho�mann &
Graeme Trousdale (eds.), The Oxford handbook of Construction Grammar, 15–31.
Oxford, UK: Oxford Univ. Press.
Kaeshammer, Miriam & Vera Demberg. 2012. German and English treebanks and lexica
for Tree-Adjoining Grammars. In Nicole�a Calzolari, Khalid Choukri, Thierry Declerck,
Mehmet Uğur Doğan, Bente Maegaard, Joseph Mariani, Asuncion Moreno, Jan Odijk &
Stelios Piperidis (eds.), Proceedings of the eighth international Conference on Language
Resources and Evaluation (LREC’12). Istanbul, Turkey: European Language Resources
Association (ELRA).
Ristad, Eric Sven. 1987. Revised General Phrase Structure Grammar. In Proceedings of
the 25th annual meeting of the Association for Computational Linguistics, 243–250.
Stanford, CA. http://www.aclweb.org/anthology/P87-1034.
Sag, Ivan A. 2012. Sign-Based Construction Grammar: An informal synopsis. In
Hans C. Boas & Ivan A. Sag (eds.), Sign-Based Construction Grammar, vol. 193 (CSLI
Lecture Notes), 69–202. Stanford, CA: CSLI Publications.
van Trijp, Remi. 2013. A comparison between Fluid Construction Grammar and
Sign-Based Construction Grammar. Constructions and Frames 5 (1). 88–116.
Uszkoreit, Hans & Stanley Peters. 1987. On some formal properties of metarules.
English. In Walter J. Savitch, Emmon Bach, William Marsh & Gila Safran-Naveh (eds.),
The formal complexity of natural language (Studies in Linguistics and Philosophy 33),
227–250. Dordrecht, The Netherlands: D. Reidel Publishing.
http://dx.doi.org/10.1007/978-94-009-3401-6_9.
Xia, Fei. 2001. Automatic grammar generation from two di�erent perspectives. University
of Pennsylvania PhD thesis.
http://faculty.washington.edu/fxia/papers_from_penn/thesis.pdf.
XTAG Research Group. 2001. A Lexicalized Tree Adjoining Grammar for English.
Tech. rep. Philadelphia, PA: Institute for Research in Cognitive Science, University of
Pennsylvania.