 
 Towards a Dependency Parser for Basque 
M. J. Aranzabe, J.M. Arriola and A. Diaz de Ilarraza, 
Ixa Group. (http://ixa.si.ehu.es) 
Department of Computer Languages and Systems 
University of the Basque Country 
P.O. box 649, E-20080 Donostia  
jibarurm@si.ehu.es 
 
 
Abstract 
We present the Dependency Parser, 
called Maxuxta, for the linguistic 
processing of Basque, which can serve 
as a representative of agglutinative 
languages that are also characterized by 
the free order of its constituents. The 
Dependency syntactic model is applied 
to establish the dependency-based 
grammatical relations between the 
components within the clause. Such a 
deep analysis is used to improve the 
output of the shallow parsing where 
syntactic structure ambiguity is not fully 
and explicitly resolved. Previous to the 
completion of the grammar for the 
dependency parsing, the design of the 
Dependency Structure-based Scheme 
had to be accomplished; we concentrated 
on issues that must be resolved by any 
practic al system that uses such models. 
This scheme was used both to the 
manual tagging of the corpus and to 
develop the parser. The manually tagged 
corpus has been used to evaluate the 
accuracy of the parser. We have 
evaluated the application of the grammar 
to corpus, measuring the linking of the 
verb with its dependents, with 
satisfactory results. 
1 Introduction 
This article describes the steps given for the 
construction of a dependency syntactic parser 
for Basque (Maxuxta ). Our dependency 
analyser follows the constraint-based approach 
advocated by Karlsson (Karlsson, 1995). It 
takes as input the information obtained in the 
shallow parsing process (Abney, 1997). The 
shallow syntax refers to POS tagging and the 
chunking rules which group sequences of 
categories into structures (chunks) to facilitate 
the dependency analysis. The dependency 
parser is considered as the module involved in 
deep parsing (see Fig. 1). In this approach, 
incomplete syntactic structures are produced 
and, thus, the process goes beyond shallow 
parsing to a deeper language analysis in an 
incremental fashion (Aduriz et al., 2004). This 
allows us to tackle unrestricted text parsing 
through descriptions that are organized in 
ordered modules, depending on the depth level 
of the analysis (see Fig. 1).  
In agglutinative languages like Basque, it is 
difficult to separate morphology from syntax. 
That is why we consider morphosyntactic 
parsing for the first phase of the shallow 
syntactic analyser. 
CG
Morphosyntactic parsing
Syntactic tagging
Chunker
Dependencies
EUSLEM 
Morpheus
Disambiguation using linguistic 
information
Disambiguation using statistical 
information
Shallow syntactic parsing
Named Entities
%
CG
PostpositionsCG
xfst
Noun and verb chainsCG
Tagging of syntactic dependenciesCG
Shallow parsing
Deep parsing
Raw data
Analysed text
 
Fig. 1. Syntactic processing for Basque. 
The dependency parser has been performed 
in order to improve the syntactic analysis 
 
achieved so far, in the sense that, apart from 
the surface structural properties, we have 
added information about deeper structures by 
expressing the relation between the head and 
the dependent in an explicit manner. 
Additionally, we have adopted solutions to 
overcome problems that have emerged in 
doing this analysis (such as discontinuous 
constituents, subordinate clauses, etc. This 
approach has been used in several projects 
(Järvinen & Tapanainen, 1998; Oflazer, 2003).  
Before carrying out the definition of the 
grammar for the parser, we established the 
syntactic tagging system in linguistic terms. 
We simultaneously have applied it to build the 
treebank for Basque (Eus3LB1) (Aduriz et al., 
2003) as well as to define the Dependency 
Grammar. The treebank would serve to 
evaluate and improve the dependency parser. 
This will enable us to check how robust our 
grammar is.  
The dependency syntactic tagging system is 
based on the framework presented in Carroll et 
al., (1998, 1999): each sentence in the corpus 
is marked up with a set of grammatical 
relations (GRs), specifying the syntactic 
dependency which holds between each head 
and its dependent(s). However, there are 
certain differences: in our system, arguments 
that are not lexicalised may appear in 
grammatical relations  (for example, the 
phonetically empty pro argument, which 
appears in the so-called pro-drop languages). 
The scheme is superficially similar to a 
syntactic depende ncy analysis in the style of 
Lin (1998). We annotate syntactically the 
Eus3LB corpus following the dependency-
based formalism. The dependencies we have 
defined constitute a hierarchy (see Fig. 2) that 
describes the theoretically and empirically 
relevant dependency tags employed in the 
analysis of the basic syntactic structures of 
Basque.  
                                                 
1This work is part of a general project 
(http://www.dlsi.ua.es/projectes/3lb) which objective is to build 
three linguistically annotated corpora with linguistic annotation 
at syntactic, semantic and pragmatic levels: Cat3LB (for 
Catalan), Cast3LB (for Spanish) (Civit & Martí, 2002) and 
Eus3LB (for Basque). The Catalan and the Spanish corpora 
include 100.000 words each, and the Basque Corpus 50.000 
words. 
This formalism is also used in the Prague 
Dependency Treebank for Czech (Hajic, 1998) 
and in NEGRA corpora for German (Brants et 
al., 2003) among others.  
 
dependant
structurally case
marked
complements
negation
linking-words
modifiers
auxiliary
others
semantics
non clausal
clausal
clausal
non
clausal
determiner
non clausal
clausal
predicative
finite
non finite
clausal
non
clausal
connector
apposition
graduator
particle
interjec.
ncsubj
nczobjncobj
ncmod
finite
non finite
detmod
xcomp_obj
xmod
xcomp_subj
cmod
ccomp_objccomp_subj
ncmod
lot
auxmod
ncpred
non finite xpred
finite
non
finite
aponcmod
apocmod
apoxmod
gradmod
prtmod
itj_out
arg_mod
meta
galdemod
ccomp_zobj
xcomp_zobj
 
Fig. 2. Dependency relations hierarchy. 
Section 2 examines the main features of the 
language involved in the analysis in terms of 
dependency relations. Taking into account 
these features, we will explain the reasons for 
choosing the dependency-based formalism. In 
section 3 we briefly describe the general 
parsing system. Section 4 explains the 
dependency relations, the implementation of 
the dependency rules and a preliminary 
evaluation. Finally, some conclusions and 
objectives for future work are presented. 
 
2 A brief description of Basque in order 
to illustrate the adequacy of the adopted 
formalism 
Basque is an agglutinative language, that is, 
for the formation of words the dictionary entry 
independently takes each of the elements 
necessary for the different functions (syntactic 
case included). More specifically, the affixes 
corresponding to the determinant, number and 
declension case are taken in this order and 
independently of each other. These elements 
appear only after the last element in the noun 
phrase. One of the main characteristics of 
 
Basque is its declension system with numerous 
cases, which differentiates it from languages 
spoken in the surrounding countries.  
At sentence level, the verb appears as the 
last element in a neutral order. That is, given 
the language typology proposed by Greenberg, 
Basque is a Subject-Object-Verb (SOV) type 
language (Laka, 1998) or a final head type 
language. However, this corresponds to the 
neutral order, but in real sentences any order of 
the sentence elements (NPs, PPs) around the 
verb is possible, that is, Basque can also be 
considered a language with free order of 
sentence constituents.  
These are the principal features that 
characterize the Basque language and, 
obviously, they have influenced us critically in 
our decision:  
 
1.  The dependency-based formalism is the one 
that could best deal with the free word order 
displayed by Basque syntax (Skut et al., 
1997). 
2.  We consider that the computational tools 
developed so far in our group facilitate 
either achieving dependency relations or 
transforming from dependency-trees to other 
modes of representation.  
3.  From our viewpoint, it is less messy to 
evaluate the relation between the elements 
that compose a sentence rather than the 
relation of elements included in parenthesis. 
4.  Dependency-based formalism provides a 
way of expressing semantic relations. 
3 Overview of the Syntactic Processing 
of Basque: from shallow parsing to deep 
parsing  
We face the creation of a robust syntactic 
analyser by implementing it in sequential rule 
layers. In most of the cases, these layers are 
realized in grammars defined by the Constraint 
Grammar formalism (Karlsson et al. , 1995; 
Tapanainen & Voutilainen, 1994). Each 
analysis layer uses the output of the previous 
layer as its input and enriches it with further 
information. Rule layers are grouped into 
modules depending on the level of depth of 
their analysis. Modularity helps to maintain 
linguistic data and makes the system easily 
customisable or reusable.  
Figure 1 shows the architecture of the 
system, for more details, see Aduriz et al., 
2004. The shallow parsing of the text begins 
with the morphosyntactic analysis and ends 
delimiting noun and verb chains. Finally, the 
deep analysis phase establishes the 
dependency-based grammatical relations 
between the components within the clause.  
The parsing system is based on finite state 
grammars. The Constraint Grammar (CG) 
formalism has been chosen in most cases 
because, on the one hand, it is suitable for 
treating unrestricted texts and, on the other 
hand, it provides a useful methodology and the 
tools to tackle morphosyntax as well as free 
order phrase components in a direct way.  
A series of grammars are implemented 
within the module of the shallow parsing 
which aim:  
1. To be useful for the disambiguatio n of 
grammatical categories, removing incorrect 
tags based on the context. 
2. To assign and disambiguate partial syntactic 
functions. 
3. To assign the corresponding tags to delimit 
verb and noun chains. 
3.1 Shallow Syntactic Analyser 
The shallow or partial parsing analyser 
produces minimal and incomplete syntactic 
structures. The output of the shallow parser, as 
stated earlier, is the main base for the 
dependency parser. The shallow syntactic 
analyser includes the following modules: 
1. The morphosyntactic analyser MORFEUS. 
The parsing process starts with the outcome 
of the morphosyntactic analyser MORFEUS 
(Alegria et al., 1996), which was created 
following a two-level morphology 
(Koskenniemi, 1983). It deals with the 
parsing of all the lexical units of a text, both 
simple words and multiword units as a 
Complex Lexical Unit (CLU).  
2. The morphosyntactic disambiguation 
module EUSLEM. From the obtained 
results, grammatical categories and lemmas 
are disambiguated. Once morphosyntactic 
disambiguation has been performed, this  
module assigns a single syntactic function to 
each word.  
 
3.  The ckunk analysis module ZATIAK. This 
module identifies verb and noun chains 
based on the information about syntactic 
functions provided by each word-form. 
Entity names and postpositional phrases are 
also determined.  
We will focus on the last step of the shallow 
analysis because it contains the more 
appropriate information to make explicit the 
dependency relations. Basically, we use the 
syntactic functions and the chunks that are 
determined in the partial analysis. 
Shallow syntactic functions 
The syntactic functions that are determined 
in the partial analysis are based on those given 
in Aduriz et al., 2000. The syntactic functions 
employed basically follow the same approach 
to syntactic tags found in ENGCG 
(Voutilainen et al., 1992), although some 
decisions and a few changes were necessary. 
There are three types of syntactic functions:  
1.  Those that represent the dependencies 
within noun chains (@CM>, @NC> etc.). 
2.  Non-dependent or main syntactic functions 
(@SUBJ, @OBJ, etc.). 
3.  Syntactic functions of the components of 
verb chains (@-FMAINV, @+FMAINV, 
etc.). 
The distinction of these three groups is 
essential when designing the rules that assign 
the function tags for verb and noun chains 
detection. 
Chunker: verb chain and noun chains 
After the morphological analysis and the 
disambiguation are performed (see Figure 1), 
we have the corpus syntactically analysed 
following the CG syntax. In this syntactic 
representation there are not phrase units. But 
on the basis of this representation, the 
identification of various kinds of phrase units 
such as verb chains and noun chains is 
reasonably straightforward.   
Verb chains  
The identification of verb chains is based on 
both the verb function tags (@+FAUXV, @-
FAUXV, @-FMAINV, @+FMAINV, etc.) and 
some particles (the negative particle, modal 
particles, etc.).  
There are two types of verb chains: 
continuous and dispersed verb chains (the 
latter consisting of three components at most). 
The following function tags have been defined: 
• %VCH: this tag is attached to a verb chain 
consisting of a single element. 
• %INIT_VCH: this tag is attached to the 
initial element of a complex verb chain. 
• %FIN_VCH: this tag is attached to the final 
element of a complex verb chain. 
The tags used to mark-up dispersed verb 
chains are: 
• %INIT_NCVCH: this tag is attached to the 
initial element of a non-continuous verb 
chain. 
• %SEC_NCVCH: this tag is attached to the 
second element of a non-continuous verb 
chain. 
• %FIN_NCVCH: this tag is attached to the 
fina l element of a non-continuous verb 
chain. 
Noun chains 
This module is based on the following 
assumption: any word having a modifier 
function tag has to be linked to some word or 
words with a main syntactic function tag. 
Moreover, a word with a main syntactic 
function tag can, by itself, constitute a phrase 
unit (for instance, noun phrases, adverbials and 
prepositional phrases). Taking into account this 
assumption, we recognise simple and 
coordinated noun chains, for which these three 
function tags have been established:  
• %NCH: this tag is attached to words with 
main syntactic function tags that constitute a 
phrase unit by themselves 
• %INIT_NCH: this tag is attached to the 
initial element of a phrase unit.  
• %FIN_NCH: this tag is attached to the final 
element of a phrase unit.  
Figure 3 shows part of the information 
obtained in the process of parsing the sentence 
Defentsako abokatuak desobedientzia 
zibilerako eskubidea aldarrikatu du epaiketan 
(The defense lawyer has claimed the right to 
civil disobedience in the  trial) with its 
corresponding chains tags.  
Let us know the some syntactic tags used in 
fig. 3: @NC>: noun complement; @CM>: 
modifier of the word carrying case in the noun 
 
chain; @-FMAINV: non finite main verb; 
@+FAUXV: finite auxiliary verb and 
@ADVL: adverbial. 
"<Defentsako>" <INIT_CAP>"   defense  
     "defentsa" N @NC>  %INIT_NCH  
"<abokatuak>"  the lawyer  
      "abokatu" N @SUBJ  %FIN_NCH  
"<desobedientzia>"                       disobedience  
   "desobedientzia" N @CM> %INIT_NCH 
"<zibilerako>"                                to civil  
       "zibil" ADJ @<NC 
"<eskubidea>"                                the right  
       "eskubide" N @OBJ %FIN_NCH 
"<aldarrikatu>"                              claimed  
   "aldarrikatu" V @-FMAINV %INIT_VCH 
"<du>"                                            has  
   "*edun" AUXV @+FAUXV %FIN_VCH   
"<epaiketan>"                                 in the trial 
        "epaiketa" N @ADVL  %NCH  
"<$.>" <PUNCT_PUNCT>" 
Fig. 3. Analysis of chains. English translation on the 
right 
3.3 Deep Syntactic Analysis  
The aim of the deep syntactic analysis is to 
make explicit the dependency relations 
between words or chunks. For this reason, we 
have designed a Dependency Grammar based 
on the Constraint Grammar Formalism. 
4 The Dependency Grammar for the 
Parser  
In this section we describe in more detail the 
dependency relations defined (see fig. 2), the 
design of the rules and the results obtained. 
The results obtained in the deep parsing of 
sample sentence will help in providing a better 
understanding of the mentioned parsing 
process. This parsing process takes as basis the 
output of the shallow parser (see fig. 3). The 
rules are implemented by means of the CG-2 
parser (www.conexor.com). 
4.1 The dependency relations 
As Lin (2003) says a dependency 
relationship (Hays, 1964; Hudson, 1984; 
Mel’cuk, 1987; Bömová et al., 2003) is an 
asymmetric binary relationship between a 
word called head (or governor, parent), and 
another word called modifier (or dependent, 
daughter). Dependency grammars represent 
sentence structures as a set of dependency 
relationships. Normally the dependency 
relationships form a tree that connects all the 
words in a sentence. A word in the sentence 
may have several modifiers, but each word 
may modify at most one word. The root of the 
dependency tree does not modify any word. It 
is also called the head of the sentence. 
For example, figure 4 describes the 
dependency structure of the example sentence. 
We use a list of tuples to represent a 
dependency tree. Each tuple represents one 
relation in the dependency tree. For example, a 
structurally case-marked complement when 
complements are nc (non-clausal, Noun 
Phrases, henceforth NP) has the following 
format: 
case : the case-mark by means of what the 
relation is established among the head and the 
modifier. 
head: the modified word head of 
NP/dependent: the modifier. In this case, the 
head of the NP. 
case-marked element within 
NP/dependent: the component of the 
dependent NP that carries the case. 
subj relationship: the label assigned to the 
dependency relationship. 
The syntactic dependencies between the 
components within the sentence are 
represented by tags starting with “&”. The 
symbols “>” and “<” attached to each 
dependency-tag represent the direction in 
which we find the sentence component whose 
dependant is the target word.  
In the example we can see that the noun 
phrase defentsako abokatuak  ‘the defense 
lawyer’ depends on the verb aldarrikatu ‘to 
claim’, which is on its right side. A post-
process will make this link explicit. 
The dependency tre e in fig 4 is represented 
by the following tuples: 
 
Modifier Cat Head Type  
Defentsako 
abokatuak 
desobedientzia 
zibilerako 
eskubidea 
aldarrikatu 
du 
epaiketan 
N 
N 
N 
ADJ 
N 
V 
Aux 
N 
abokatuak  
aldarrikatu  
eskubidea  
desobedientzia 
aldarrikatu 
 
aldarrikatu 
aldarrikatu 
&NCMOD> 
&NCSUBJ> 
&NCMOD> 
&<NCMOD 
&NCOBJ> 
 
&<AUXMOD 
&<NCMOD 
 
4.2 The dependency grammar rules  
The grammar consists of 255 rules that have 
been defined and distributed in the following 
way: 
 
complements modifiers 
nc2 cc3 det nc  cm4 
others 
62 11 19 124 20 19 
 
These rules were formulated, implemented, 
and tested using a part of the manually 
disambiguated corpus (24.000 words). For the 
moment, part of the rest of the corpus was used 
for testing.  
For more details of the rules, we describe 
some example s that illustrate how dependency 
rules can be written to define different types of 
linguistic relations. 
 
1. Verb-subject dependency 
The following rule defines a verb-subject 
dependency relation between 2 words 
aldarrikatu (claimed) and abokatuak   (lawyer) 
of the sentence in the previous example:  
  
 MAP (&NCSUBJ>) TARGET (NOUN)  
   IF (0 (ERG) + (@SUBJ) +(%FIN_NCH)) 
      (*1(@-FMAINV) + (%INIT_VCH)  
       BARRIER (PUNCT_PUNCT)); 
 
The rule assigned the ncsubj tag to the noun 
abokatuak (lawyer) if the following conditions 
are satisfied: a) the noun is declined in ergative 
case; besides, it has assigned the @SUBJ 
syntactic function and, it is the last word of a 
noun chain; b) it has a non-finite main verb 
everywhere on its right before the punctuation 
mark. 
                                                 
2 nc: non-clausal complement or modifier 
3 cc:clausal complement 
4 cm: clausal modifier 
 
2. Subordinate clause dependency 
The following rule defines a complement 
subordinate clause dependency relation 
between a subordinate verb and a main verb. 
We illustrate this rule by means of an example 
in which the word egoten (usually stayed) is 
the verb of the complement subordinate clause 
linked to esan (told): 
 
Example: Lehenago aitona egoten zela ni 
EGOTEN naizen tokian esan dit amonak5. 
 
 MAP(&CCOMP>>)TARGET (V)  
 IF(0(@-FMAINV)+ (%INIT_VCH)) 
(1(@+FAUXV_SUB)+ (%FIN_VCH)); 
 
The rule assigned the CCOMP tag to the 
verb egoten  (usually stayed) if the following 
conditions are satisfied: a) the verb is a non-
finite main verb and, it’s the first word-form of 
a verb chain; b) it has an auxiliary verb on its 
immediate right-side which has assigned the 
complement tag and appears as the last part of 
the verb chain.  
 
3. Infinitive control 
The following rule defines that in the 
sentence Jonek Miren etortzea nahi du. (John 
wants to come Mary), etortzea (infinitive 
subordinate clause with object function, "to 
come") is controled by the main verb nahi  ("to 
want"). Taking into account, that etortzea  is 
the controlled object of nahi, if there is another 
non-infinitive object Miren; then we will 
assign to it the subject dependency relation to 
the infinitive verb ("to come").   
  
                                                 
5 My grandmother told me my grandfather 
usually stayed  where I am now 
epaiketan Defentsako abokatuak desobedientzia  zibilerako eskubidea aldarrikatu du 
Fig.4. Dependency tree 
 
MAP (&NCSUBJ>) TARGET (NOUN)  
IF (0 (ABS) + (@SUBJ) OR (@OBJ)  + (%NCH))  
    (1(@-FMAINV_SUB_@OBJ) ) (2 VTRANS_ -FV )); 
  
4.3 Evaluation 
The system has been manually tested on a 
corpus of newspaper articles (included in 
Eus3LB), containing 302 sentences (3266 
words).  
We have evaluated the precision (correctly 
selected dependent / number of dependant 
returned) and the recall (correctly selected 
dependent / actual dependent in the sentence) 
of the subject (including coordinated subjects), 
and modifier dependency of verbs. For subject, 
precision and recall were respectively 67% and  
69 %, while the figures for verb modifiers were 
73 % and   95%. 
We have detected two main  reasons for 
explaining these figures: 1) the analysis 
strategy is limited because we cannot make use 
of semantic or contextual information for 
resolving uncertainties at an early level; 2) 
errors in previous steps. These errors can be a) 
due either to an incorrect assignment of POS to 
word-forms or to the syncretism of case marks 
(@SUBJ, @OBJ); b) the presence of non-
known word-forms that increases the number 
of possible analysis. At this moment, the head 
and dependent slot fillers are, in all cases, the 
base forms of single head words, so for 
example, ‘multi-component’ heads, such as 
names, are reduced to a single word; thus the 
slot filler corresponding to Xabier Arzallus 
would be Arzallus.  
5 Conclusions 
We have presented the application of the 
dependency grammar parser for the processing 
of Basque, which can serve as a representative 
of agglutinative languages with free order of 
constituents.  
We have shown how dependency grammar 
approach provides a good solution for deeper 
syntactic analysis, being at this moment the 
best alternative for morphologically complex 
languages.  
We have also evaluated the application of 
the grammar to corpus, measuring the linking 
of the verb with its dependents, with 
satisfactory results. However, the development 
of a full dependency syntactic analyser is still a 
matter of research.  For instance, all kinds of 
constructions without a clear syntactic head are 
difficult to analyse: ellipses, sentences without 
a verb (e.g., copula -less predicative), and 
coordination. All these aspects have been 
treated in our manually annotated Corpus; our 
efforts now are oriented to deal with them 
automatically. 
 
6 Acnowledgments  
This research is supported by the University 
of the Basque Country (9/UPV00141.226-
14601/2002), the Ministry of Industry of the 
Basque Government (project XUXENG, 
OD02UN52). 

References  
Abney S. P. 1997. Part-of-speech tagging and 
partial parsing. S. Young and G. Bloothooft, 
editors,  Corpus -Based Methods in Language 
and Speech Processing, Kluwer, Dordrecht. 
Aduriz I., Aranzabe M.J., Arriola J.M.,  Díaz 
de Ilarraza A., Gojenola K., Oronoz M., Uría 
L. 2004. A Cascaded Syntactic Analyser for 
Basque. In Gelbukh, A (ed.) Computational 
Linguistics and Intelligent Text Processing. 
SpringerLNCS 2945.  
Aduriz I., Aranzabe M.J., Arriola J.M., Atutxa 
A., Díaz de Ilarraza A., Garmendia A., 
Oronoz M. 2003. Construction of a Basque 
Dependency Treebank. Proceedings of the 
Second Workshop on Treebanks and 
Linguistic Theories "TLT 2003", (J. Nivre 
and E. Hinrichs eds.), Växjö University 
Press. V äxjö, Suecia   
Aduriz I., Arriola J.M., Artola X., Diaz de 
Illarraza A., Gojenola K., Maritxalar M. 
2000. Euskararako Murriztapen Gramatika: 
mapaketak, erregela morfosintaktikoak eta 
sintaktikoak. UPV/EHU/LSI/TR 12-2000.  
Alegria I., Artola X., Sarasola K., Urkia M. 
1996. Automatic morphological analysis of 
Basque. Literary & Lin guistic Computing 
Vol. 11, No. 4, 193-203. Oxford University 
Press. Oxford.  
Bömová , A., Haji¡c, J., Hajicov´a, E., 
Hladk´a, B. 2003. The Prague 
DependencyTreebank: A Three level 
Annotation Scenario. In Abeillé (ed.) 
Treebanks Building and Using Parsed 
Corpora, Book Series: TEXT, SPEECH 
AND LANGUAGE TECHNOLOGY : 
Volume 20 Kluwer Academic Publisher, 
Dordrecht. 
Brants T., Skut W. & Uszkoreit H. 2003 
"Syntactic Annotation of a German Newspa-  
per Corpus”. In Abeillé (ed.) Treebanks 
Building and Using Parsed Corpora, Book 
Series: TEXT, SPEECH AND LANGUAGE 
TECHNOLOGY : Volume 20 Kluwer 
Academic Publisher, Dordrecht. 
Carroll J., Briscoe E., Sanfilippo A. 1998. 
Parser evaluation: a survey and a new 
proposal. Proceedings of the 1st 
International Conference on Language 
Resources and Evaluation, 447-454. 
Granada, Spain.  
Carroll J., Minnen G., Briscoe T. 1999. Corpus 
Annotation for Parser Evaluation. 
Proceedings of Workshop on Linguistically 
Interpretated Corpora, EACL´99. Bergen. 
Civit M. & Martí M. 2002. Design Principles 
for a Spanish Treebank. Proceedings of The 
Treebank and Linguistic Theories 
(TLT2002). Sozopol, Bulgaria. 
Hays, D. 1964. Dependency theory: a 
formalism and some observations. 
Language40, p. 511–525. 
Hajic J. 1998. Building a Syntactically 
Annotated Corpus: The Prague Dependency 
Treebank. In  Issues of Valency and 
Meaning, 106-132. Karolinum, Praha. 
Hudson, R. 1984. Word Grammar. Oxford, 
England: Basil Blackwell PublishersLimited. 
Järvinen T. and Tapanainen P, 1998. Towards 
an implementable dependency grammar. In 
Proceedings of the Workshop "Processing of 
Dependency-Based Grammars", (eds.) 
Sylvain Kahane and Alain Polguère, 
Université de Montréal, Quebec, Canada, 
15th August 1998, pp. 1-10. 
Karlsson F., Voutilainen A., Heikkila J., 
Anttila A. 1995. Constraint Grammar: a 
Language-Independent System for Parsing 
Unrestricted Text. Mouton de Gruyter. 
Koskenniemi K 1983. Two-level Morphology: 
A general Computational Model for Word-
Form Recognition and Production. 
University of Helsinki, Department of 
General Linguistics. Publications 11.  
Laka, I. 1998. A Brief Grammar of Euskara, 
the Basque Language. HTML document. 
http://www.ehu.es/grammar. Office of the 
Vice-Dean for the Basque Language. 
University of the Basque Country. 
Lin D. 1998. A Dependency-based Method for 
Evaluating Broad-Coverage Parsers. Natural 
Language Engineering.  
Lin D. 2003. "Dependency-based evaluation of 
MINIPAR" in Building and Using 
syntactically annotated corpora, Abeillé, A. 
Ed. Kluwer, Dordrecht 
Mel’cuk, I. A. 1987. Dependency syntax: 
theory and practice. Albany: StateUniversity 
of New York Press. 
Oflazer K. 2003. Dependency Parsing with an 
Extended Finite-State Approach. ACL 
Journal of Computational Linguistics, Vol. 
29, nº4. 
Skut W., Krenn B., Brants T., Uszkoreit H. 
1997. An Annotation Scheme for Free Word 
Order Languages. In Proceedings of the 
Fifth Conference on Applied Natural 
Language Processing (ANLP-97). 
Washington, DC, USA. 
Tapanainen P. and Voutilainen A. 1994 
Tagging Accurately-Don´t guess if you know. 
In Proceedings of the 4th Conference on  
Applied Natural Language Processing, 
Washington. 
Voutilainen A., Heikkilä J. and Anttila A. 
1992. Constraint Grammar of English. A 
Performance-Oriented Introduction. 
Publications of Department of General 
Linguistics, University of Helsinki, No. 21, 
Helsinki. 
