fql016 187..198


Functional Disambiguation Based
on Syntactic Structures
............................................................................................................................................................

Octavio Santana Suárez, José Rafael Pérez Aguiar,

Luis Losada Garcı́a, and

Francisco Javier Carreras Riudavets

University of Las Palmas de Gran Canaria
.......................................................................................................................................

Abstract
This article presents a disambiguation method which diminishes the functional
combinations of the words of a sentence taking into account the context in which
they appear. This process is built in two phases: the first phase is based on the
local syntactic structures of the Spanish language and reaches an average yield of
87%. The second one is supported by syntactic tree representation and pushes the
results up to an approximate high end of 96%. This process constitutes the
starting point towards an automated syntactic analysis.

.................................................................................................................................................................................

1 Introduction

In the Spanish language, there are a considerable
number of words that can play different gramma-
tical functions, and therefore a text analysis would
produce an enormous amount of combinations
unless the function of each word within the
context where it appears is considered. Functional
disambiguation consists of the elimination of the
results that do not answer to their function within
the text.

This article presents a method of functional
disambiguation this method reduces the size
of the answer through a two-step treatment of a
morphological processor. In the first stage, a
functional disambiguation based on local syntactic
structures is applied; here the grammatical functions
that invalidate the neighbouring environment of
every word within the sentence are discarded. In the
second stage, the functional disambiguation is
performed; at this point the combinations of gram-
matical functions of the sentence that prevent the
generation of syntactic representation trees valid for
the whole sentence are discarded.

2 Basic Syntactic Structures and
Functional Pairs

In the Spanish language, there are basic structures
that repeat and combine over and over among
themselves in order to give way to the sentences of
the discourse. The composition of these structures
defines the pairs of grammatical functions that
appear in a sentence—within these local-type
structures. When a local-type study is to be
performed, the null symbol is included both at the
beginning and at the end of every structure.

The functional behaviours of the following need
to be considered: noun, adjective, demonstrative
adjective, possessive adjective, adverb, personal
pronoun, relative pronoun, remaining pronouns,
article, preposition, conjunction, coordinating con-
junction and contraction.

Some categories are disclosed because they show
function and position differences in the syntactic
structures. Among adjectives it is possible to
distinguish the possessive ones from the demon-
strative ones; the possessive adjectives that can
appear before, after and in both positions in relation

Correspondence:

Francisco Javier Carreras

Riudavets,

Departamento de

Informática y Sistemas,

Edificio de Informática y

Matemáticas, Campus

Universitario de Tafira,

Universidad de Las Palmas

de Gran Canaria, 35017 Las

Palmas de Gran Canaria,

Las Palmas, Spain.

E-mail:

fcarreras@dis.ulpgc.es

Literary and Linguistic Computing, Vol. 21, No. 2, 2006. � The Author 2006. Published by Oxford University Press on
behalf of ALLC and ACH. All rights reserved. For Permissions, please email: journals.permissions@oxfordjournals.org

187

doi:10.1093/llc/fql016



to the nominal head which they complement can be
separated. Among pronouns we can with distin-
guish the demonstrative adjectives—with adjective
function—from the personal pronouns—the
unstressed and the tonic pronouns are identified
separately—and from the relative pronouns;
the remaining group of pronouns are considered
under the denomination of other pronouns. The
coordinating conjunctions will be taken into
account in a special fashion because they are used
to link formal structures of the same syntactic
level—all of them are included under the denomi-
nation of a conjunction. The personal forms of
the infinitive, gerund and participle can also be
distinguished from each other.

Among the contracted forms, a combination of a
preposition and a determiner and, sometimes, a
combination of three elements will be considered.
The punctuation marks are also considered, differ-
entiating between a comma and a semicolon.

2.1 Homogeneous noun phrase
The homogeneous noun phrase has the following
basic structure: null þ determiner þ nominal head þ
adjacencies þ null. The determiner may be: an
article, a possessive adjective or a demonstrative
adjective. The nominal head is formed by a noun.
The adjacency may be: an adjective, the de
preposition followed by a phrase (prepositional
complement of the noun) or a noun (apposition).
The determiner, the nominal head and the adja-
cency must agree in gender and number.

This structure may exhibit certain variations in
relation to the presence and position of their
elements. The nominal head will always be present;
however, the determiner and the adjacency may not
appear. Some adjacencies—adjective—may precede
the nominal head and, sometimes, the determiner—
possessive adjective—may follow the nominal head.

Table 1 shows the configurations formed by
consecutive pairs are shown.

2.2 Heterogeneous noun phrase
Heterogeneous noun phrases are combinations
of the homogeneous ones: homogeneous noun
phrase þ connector þ homogeneous noun phrase.

The connectors are conjunctions, from the gram-
matical point of view, and from the graphical point
of view they are realized by the comma (,). The new
combinations of symbols that appear are listed in
Table 2.

2.3 Substitute noun phrase
The substitute noun phrase appears when the
nominal head is realized by a category different
from the noun: the pronoun, the adjectives and the
infinitives preceded by a determiner. With respect
to the homogeneous noun phrase, they appear as
new pairs of functional categories as a result of
substituting the noun that forms the head by a
pronoun (tonic personal pronoun, pronoun of
relative preceded by an article or other pronoun),
an adjective (preceded by an article or a demon-
strative adjective) or an infinitive (preceded by an
article, possessive adjective or demonstrative adjec-
tive). The head continues to conform with the
determiner and the adjacencies with regard to
gender and number.

2.4 Verb
The simple verbal forms are constituted by one
verb in the active mode with the basic structure:
null þ simple verbal form þ null. A simple verbal
form may be a personal verbal form or an infinitive.

Table 1 Pairs of symbols that form homogeneous noun

phrase

Followed by null determiner nominal head adjacency

null no yes yes yes

determiner yes no yes yes

nominal head yes yes no yes

adjacency yes yes yes yes

Table 2 Pairs of symbols that form heterogeneous noun

phrase

Preceded by connector Followed by connector

determiner yes yes

nominal head yes yes

adjacency yes yes

O. S. Suárez et al.

188 Literary and Linguistic Computing, Vol. 21, No. 2, 2006



A complex verbal form has the following basic
structures: null þ auxiliary þ impersonal form þ null
and null þ proclitic þ personal form þ null. The first
structure includes the already understood auxiliary
followed by a participle, compound tenses, passive
voice, the auxiliary of direct action followed by
an infinitive and an auxiliary verb followed by
a gerund. The second structure represents an
unstressed personal pronoun followed by a simple
or a compound personal verbal form.

In addition, there are two special cases that must
be treated: null þ indirect incidence auxiliary þ
conjunction þ infinitive þ null (because the only
acceptable conjunction is que) and null þ indirect
incidence auxiliary þ preposition þ infinitive þ null
(because the acceptable prepositions are a, de, en
and por).

Finally, the existence of multiple verbal heads
have to be taken into account: verb þ connector þ
verb (the connectors are again conjunctions from
the grammatical point of view and the comma from
the graphical point of view).

2.5 Prepositional phrase
The prepositional phrase comprises a preposition
plus a noun phrase; the pairs of contributing
functional categories are combinations with the
preposition preceded by null or followed by a
determiner, a nominal head, an adjacency and
another preposition (in the case of double prep-
osition, the first one has to be a or hasta).

2.6 Adjectival phrase
The simple adjectival phrase, which exists only with
copulative verbs, acts as an attribute, and it is
formed by an adjective in the basic structure:
null þ adjectival phrase head þ null. The adjectival
phrase head is an adjective.

In the multiple adjectival phrases, adjectives may
appear linked by connectors: null þ adjectival phrase
head þ connector þ adjectival phrase head þ null;
the connector is a coordinating conjunction or a
comma.

2.7 Adverbial phrase
The syntactic structures of the adverbial phrase are:
null þ adverb þ null, null þ adverb þ adverb þ null,

null þ adverb þ prepositional phrase þ null, null þ
adverb þ nominal phrase þ null and null þ adverbial
phrase þ null.

The pairs of contributing functional categories
are combinations where the adverb is preceded
by null and a preposition (in the case of an
adverbial phrase), or followed by another adverb,
a preposition, a determiner, a nominal head or an
adjacency.

The adverb may appear, in some cases, adjacent
to an adjective within a noun phrase; because
the combination: definite article þ adverb must be
added.

2.8 Linkage among several structures
The basic structures are combined in order to
generate structures of a larger size. In many cases,
it is not necessary to use linking particles, but
in others it is. When the structures to be linked
are clauses, it is necessary to use a linking element.
For this reason the pairs: null þ linking element
and linking element þ null are added; the linking
element can be a conjunction, a comma or a
semicolon.

3 Local Functional Disambiguation

Local functional disambiguation starts from the
following data:

(1) The allowed set of functional behaviour S—
referred to in Section 2.

(2) The set of pairs P of symbols of the
form a þ b, where a and b belong to S,
that can exist in the local structures in the
Spanish language—they have been presented
in Section 2.

(3) A set of combinations of functional categories
that are not allowed.

Due to the existence of rules of the
form null þ category and category þ null—
beginnings and endings of local structures—
disallowed combinations may occur; to avoid
this, a set of functional structures that are not
prohibited has to be defined (Table 3).

Functional Disambiguation Based on Syntactic Structures

Literary and Linguistic Computing, Vol. 21, No. 2, 2006 189



(4) Starting from the words that produce
the combination, a set of special cases is
defined:

� When there is more than one verbal form
without a link relation, the option is
disregarded.

� If there is a determiner but no adjacencies,
the ambiguity between the adjective and
the noun is resolved in favour of the noun
because it is the head of the noun phrase.

� In the case of ambiguity between adjective
and participle, the term favoured is the
adjective if there are no auxiliary verbs—
haber and ser.

� In order to avoid problems of cacophony,
the concordance of the de article with the
nominal head which it precedes is not
necessary.

� Before mı́, ti or sı́, only a preposition may
appear.

� After a question or exclamation mark, qué
will be another pronoun.

� After a verb or an adverb, que will be a
conjunction; after el, la, las, lo, los it will be
a relative pronoun; after a comma it will be
a pronoun or a conjunction.

� Before a noun, de will be a preposition.
� The word no acts only as a noun after el

or un.
� The words sobre and muy do not have

the value of a noun before another noun.

The following steps are executed:

(1) Processing of the morphological analysis of
the sentence and getting a set of potential
functional combinations.

(2) Examining all strings in groups of three
elements in order to accept or reject
the central element. Given the sequence of
functions a þ b þ c, b is accepted if and
only if it is given any one of the following
conditions:

(1) {a þ b} and {b þ c} belong to P
(2) {null þ b} and {b þ c} belong to P
(3) {a þ b} and {b þ null} belong to P
(4) {null þ b} and {b þ null} belong to P

(3) Of the sequences not rejected, those contain-
ing any prohibited functional elements are
eliminated.

(4) The remaining combinations also include
some that fit the special cases.

4 Structural Ambiguities

Starting from a combination of functional behav-
iours of the words of a sentence, it is possible to get
more than one tree for the analysis when applying
the Spanish grammar considered here—it is forma-
lized with more than 400 rules; such multiple results
denote structural ambiguity.

The existence of more than one rule with the
same symbol or combination of symbols on the
right side is what is denominated direct structural
ambiguity; the grammar used here comprises more
than 60 direct structural ambiguities that cover
about 240 rules. The direct structural ambiguities
led to primary conflicts. There are cases of real
ambiguities that can lead to more than one valid
interpretation of a sentence.

Table 3 Prohibited combinations

Prohibited functional combinations

preceding possessive adjective þ preposition

preceding or after possessive adjective þ preceding

possessive adjective

definite article þ preposition

comma þ conjunction þ punctuation

conjunction þ null

null þ demonstrative adjective þ personal verbal form

pronoun þ infinitive

unstressed personal pronoun þ adjective

unstressed personal pronoun þ adverb

unstressed personal pronoun þ determined article

unstressed personal pronoun þ conjunction

unstressed personal pronoun þ preposition

unstressed personal pronoun þ pronoun

unstressed personal pronoun þ noun

punctuation þ conjunction þ comma

noun þ participle

. . .

O. S. Suárez et al.

190 Literary and Linguistic Computing, Vol. 21, No. 2, 2006



5 Solving Primary Conflicts

In the following paragraphs several proposals
for producing rules to resolve conflicts will be
considered; the superposition of these rules will
produce the removal of the non-acceptable trees of
analysis. In some cases, the rules may be applied the
very moment a new symbol is added during the
process of analysis, i.e. when the rules depend upon
the symbols of the lower levels; in other cases it
would be necessary to wait until the completion of
the tree.

5.1 Ambiguities and necessary words
For some of the complements it is not possible to
use all the words of a given functional behaviour: in
this sense any pronoun neither originates a direct
object nor does any preposition of a prepositional
phrase give way to an indirect object.

Rule: Necessary Words
Let S be a non-terminal symbol generated starting
from an intermediate symbol IS and let PN(S) be
the set of words necessary for S, the S symbol will
be accepted if and only if it is found to belong to the
set PN(S) among the words generated by IS.

5.1.1 Prepositional Phrases

Various structures can be generated from IS ¼
prepositional phrase; however, the allowed preposi-
tion are not the same for all the structures.

The direct object, for example, only takes a and
the indirect object takes a or para. In this sense,
conflicts are eliminated in some cases and in the
remaining ones the conflicts are diminished.

In the case of concatenating prepositions the same
rule is applicable, and is applied to the second

preposition. When a contraction appears, the same
considerations for the preposition are to be applied.

There are words that, in general, are not
recognized as prepositions, but they have similar
functional behaviours; they are the so-called imper-
fect prepositions.

5.1.2 Unstressed personal pronouns

Various structures can be generated starting from
IS ¼ unstressed personal pronoun; however, the
allowed pronouns are not the same for all the
structures (Table 4).

5.1.3 Other categories

Other categories used in resolving conflicts are
shown in Table 5.

5.2 Ambiguities and symbols not
allowed
If starting from an unstressed personal pronoun, a
substitutive noun phrase is generated. It should not
give place to a direct object because such pronouns

Table 5 Other categories in the resolution of conflicts

S ¼ Structure IS ¼ Category NW(S) ¼ Necessary Words

adjacency adverb como, más, menos, no, todo/a

adjacency relative pronoun cuyo/a, cuyos/as, que

subordinate connector adverb apenas, como, conforme, cuanto, donde, mientras, siempre, tal, tan

subordinate connector conjunction aunque, con que, cuando, cuantos/as, para, porque, que, si

comparative construction adverb ası́, como

adjectival group adverb como

. . . . . . . . .

Table 4 Unstressed personal pronouns in the resolution

of conflicts

S ¼ Structure NW(S) ¼ Necessary Words

direct object la/s, lo/s, me, nos, os, se, te

indirect object la/s, le/s, lo/s, me, nos, os, se, te

attribute lo

morpheme of passive

construction

se

morpheme of impersonal

construction

se

. . . . . .

Functional Disambiguation Based on Syntactic Structures

Literary and Linguistic Computing, Vol. 21, No. 2, 2006 191



may have a direct object function only when they
are preceded by a preposition.

Rule: Non-Allowed Symbols
Let S be a non-terminal symbol, generated from an
IS symbol and let NAS(S,IS) be a set of symbols
generating IS, and catalogued as non-allowed
generators of S. Then, S generated by IS will be
rejected if IS has been generated by means of some
symbol of the set NAS(S,IS) (Table 6).

5.3 Ambiguities and related symbols
Some symbols may not appear without the existence
of other symbols in the same tree of analysis.

Rule: Necessary Symbols
The symbol S is added to the tree of analysis only if
it exists as the NS(S) symbol (Table 7).

In order to reduce the appearance of direct
objects erroneously recognized as such, it is
advisable to take into account the need for having
a transitive verb. Taking also into account that
copulative verbs are of the intransitive type the

possibility of confusing an attribute with a direct
object is, hence, reduced.

Rule: Necessary Symbols with Condition
The S symbol is added to the tree of analysis if and
only if it exists as an NS symbol that complies with
the condition C(S, NS).

5.4 Ambiguities and incompatible
symbols
The differentiation between ambiguities and incom-
patible symbols is based on the non-existence of a
symbol and not on its existence; in order to accept
the intransitive sentence symbol the direct object
symbol must not exist. Every tree that includes
incompatible symbols is rejected—with the excep-
tion of compound sentences which consist of several
predicates.

Rule: Incompatible Symbols
The symbol S is added to the analysis tree if and
only if the InS(S) symbol does not exist (Table 8).

5.5 Concordances
Among the different structures which constitute a
sentence there are mandatory requirements regard-
ing the concordance of certain characteristics.

Rule: Concordances
If S1 and S2 are the symbols of an analysis tree,
this tree is accepted if and only if there is
concordance between the set of definite character-
istics for these symbols, CSD(S1, S2).

5.5.1 Description of cases

Concordances that have been checked during the
process of local functional disambiguation should
be verified again because two elements that
appear in the same phrase must concord and in

Table 6 Non-Allowed Symbols (NAS)

S ¼ Symbol IS ¼ Intermediate Symbol NAS(S,IS) ¼ Non-Allowed Symbol

direct object noun phrase infinitive

indirect object noun phrase unstressed personal pronoun

subject noun phrase tonic personal pronoun

noun phrase nominal head adjective

. . . . . . . . .

Table 7 Necessary Symbols (NS(s))

S—New Symbol NS(S)—Necessary Symbol

attribute copulative verbal head

copulative verbal head attribute

passive verbal head passive auxiliary

direct object verbal head

attributive sentence attribute

supplement sentence supplement

intransitive sentence verbal head

passive sentence passive verbal head

transitive sentence direct object

. . . . . .

O. S. Suárez et al.

192 Literary and Linguistic Computing, Vol. 21, No. 2, 2006



the local analysis it might be that this was not
the case because the union of local structures was
assumed.

Concordance between subject and verbal head:
in sentences with verbs in a personal form there
must be concordance in number and person with
the head of the subject structure.

CSD(subject, verbal head ) ¼ {number, person}
CSD(subject, passive verbal head ) ¼ {number,
person}

CSD(subject, copulative verbal head ) ¼
{number, person}

Besides the concordance between subject and
predicate, the concordance between the next
descendants of the predicate must be accom-
plished—always in gender and number:

Adjacency with nominal head.
Direct object with direct object.
Indirect object with indirect object.
Objective predicative with direct object.
Subjective predicative with verbal head.
Subjective predicative with subject.
Determiner with nominal head.

5.6 Semantic information
The analysis of the semantic content of the words
leads to the elimination of ambiguities—starting
from the information given through ideological
dictionaries: to generate the symbol circumstantial
complement of tense among the words that form it,
there must be something that gives information on
tense or moment.

Rule: Necessary Semantic
If WS is a set of words that are joined to build the S
symbol and if IMS(S) is the set of ideological
meanings associated with the S symbol, then S is
rejected if there is no word in WS such that the
ideological analysis belongs to IMS(S).

To improve the efficiency of the automation of
the disambiguation process, it should be easy to
create a disposition containing the words with the
necessary semantics for all the symbols.

To avoid taking into account words with seman-
tics that must not directly intervene with the symbol,
the set of WS words will be formed only with words
at the highest level of the representation tree for
the symbol—if such a level contains only irrelevant
words, determiners, connectors and prepositions,
the action is to go to lower levels until relevant words
are found.

5.6.1 Ideological relationships and symbols

There are words whose semantic content prohibits
the generation of a given symbol from another: a
homogeneous noun phrase cannot generate a

Table 8 Incompatible symbols

S—New Symbol InS(S)—Incompatible Symbol

attribute attribute

attribute morpheme of passive

attribute direct object

attribute indirect object

attribute objective predicative

attribute subjective predicative

attribute supplement

agent complement supplement

morpheme of impersonal morpheme of impersonal

morpheme of impersonal morpheme of passive

morpheme of impersonal morpheme of half voice

morpheme of passive morpheme of passive

morpheme of passive morpheme of half voice

morpheme of half voice morpheme of half voice

verbal head verbal head

direct object attribute

direct object morpheme of passive

indirect object attribute

indirect object morpheme of passive

attributive sentence passive verbal head

supplement sentence passive verbal head

supplement sentence direct object

intransitive sentence attribute

intransitive sentence direct object

intransitive sentence supplement

transitive sentence attribute

transitive sentence passive verbal head

objective predicative attribute

objective predicative objective predicative

objective predicative subjective predicative

subjective predicative attribute

subjective predicative morpheme of impersonal

subjective predicative subjective predicative

subject morpheme of impersonal

supplement attribute

supplement agent complement

supplement supplement

. . . . . .

Functional Disambiguation Based on Syntactic Structures

Literary and Linguistic Computing, Vol. 21, No. 2, 2006 193



direct object if the head is a person—ideological
information.

Rule: Incompatible Semantic
If WS is a set of words that are joined to form
the S symbol starting from the IS symbol and if
SIMR(S, IS) is the set of ideological meanings that
produce the rejection of the S symbol, generated
from IS, then S is rejected if there is some word
exists in WS such that its ideological analysis
belongs to SIMR(S, IS).

5.6.2 Ideological relationships among symbols

Symbols can be used as relationships of the
ideological type between the subject head and the
verbal head: if the verbal head implies an action
and is in active form, the subject should constitute
a living being.

Rule: Ideological Relations among Symbols
If S1 and S2 are symbols of an analysis tree, this tree
is accepted if and only if the ideological concor-
dance is satisfied in the set of definite characteristics
for these symbols, IRS(S1, S2).

5.7 Special cases
Special circumstances are applied when the previous
methods do not solve the problem.

5.7.1 Clauses

The clause, whether coordinate or subordinate,
must have a connector, either a subordinator or
a coordinator, inserted before or after the clause.
Similarly, any type of sentence that allows any of
these clauses should comply with the same condi-
tions. The main clause and the subordinate clause
are different in the way they are joined to the
remainder of the sentence. The subordinate clause
of an infinitive should have an infinitive as the
verbal head.

5.7.2 Interrogative sentences and exclamatory
sentences

The interrogative and exclamatory sentences are
differentiated by the punctuation marks that delimit
them.

5.7.3 Double direct object

A double direct object-left dislocation is easily
recognizable by the following characteristics: (1) the
two elements are found together, (2) the first one
is found at the beginning of the sentence, (3) the
second one is a pronominal clitic and (4) there must
be concordance with regard to gender and number
between the two corresponding heads.

Rule: Double Direct Object
If S is a root symbol that covers the whole sentence,
and if two direct object symbols appear, then S is
accepted if and only if the direct objects are adjacent,
are followed by a verbal head and the second direct
object is realized by an unstressed personal pronoun.

It must be taken into account that in compound
sentences there may be two direct objects for each
personal verbal form.

5.7.4 Elimination of options according to the
position of the determiners

Only the following symbols: demonstrative adjec-
tive, definite article and other pronoun appear
before the nominal head. The possessive adjectives
can be divided into those that precede the nominal
head—mi, mis, tu, tus, su and sus—and those that
come after the nominal head—mı́o, mı́a, mı́os, mı́as,
tuyo, tuya, tuyos, tuyas, suyo, suya, suyos and suyas—
and those that can appear both before as well as
after the nominal head—nuestro, nuestra, nuestros,
nuestras, vuestro, vuestra, vuestros and vuestras.

Rule: Post-head Determiners
If S is a symbol that belongs to the group of
noun phrases and is generated starting from
a sequence of symbols where the nominal head þ
determiner sequence appears, S will be accepted if
and only if the determiner symbol is found in
the PSS (Post-head Symbol Set) group of terminal
symbols that can follow the nominal head.

Rule: Pre-head Determiners

If S is the symbol for the adjective phrase that is
generated starting from an adjective symbol, S will
not be accepted if it is found after a determiner
symbol from the BSS (Before Symbol Set) group of
the terminal symbols that cannot follow the
nominal head.

O. S. Suárez et al.

194 Literary and Linguistic Computing, Vol. 21, No. 2, 2006



5.7.5 Connectors

There are a number of combinations of words
that give place to conjunctive conjunctions: a
consecuencia, a distinción de, a fin de, a fin de que,
a lo que parece, a medida que, a menos que, a pesar, a
pesar de, ahora bien, ahora que, al menos, al objeto de,
al objeto de que, al parecer, al paso que, antes bien, ası́
como, ası́ es que, ası́ pués ası́ y todo, aún cuando, etc.

5.7.6 Other cases

There are situations in which ambiguities can be
resolved starting from considerations regarding
the words, grammatical categories and intervening
objects.

6 Resolutions of Other Conflicts

There are rules that, without being directly applied
to a given primary conflict, serve to eliminate
ambiguities.

6.1 Symbols that cannot cover the
whole sentence
The structure of a sentence should include a subject
and a predicate, or possibly only a predicate. The
object of analysis cannot comprise the subject
symbol by itself.

The main clause and the subordinate clause are
symbols that have been defined to generate the
analysis of compound sentences; it is for this reason
that a main clause symbol has the same structure
as a sentence symbol so that the sentence symbols
need not cover the whole sentence.

Rule: Total Symbols
The S non-terminal symbol that covers all sequences
to be analyzed is accepted if and only if it is found
among the symbols of the set TSS (Total Symbols Set)
that have allowed the covering of whole sentences.

6.2 Verbal periphrasis
The verbal periphrasis formed by more than two
elements generates a complex verbal form only
in specific cases: acabar de þ infinitive, deber
de þ infinitive, dejar de þ infinitive, echarse
a þ infinitive, empezar a þ infinitive, estar para þ
infinitive, explotar a þ infinitive, haber

de þ infinitive, haber que þ infinitive, ir
a þ infinitive, llegar a þ infinitive, ponerse
a þ infinitive, romper a þ infinitive, tener
que þ infinitive, venir a þ infinitive, volver
a þ infinitive, etc.

6.3 Considerations regarding predicate
symbol generation
The rules that define the structures of the predicate
are given through the combinations of elements
that can appear in it. In a formal definition of
structural type it would be necessary to indicate
all the possible combinations; because the place-
ment of the majority of the elements is free,
the number of possible structures for the predicate
would be enormous; therefore it has been decided
to permit all combinations and to prohibit those
that are not possible—Table 8 shows pairs of
incompatible symbols in the same predicate.

The generation of a predicate symbol should be
rejected either when some of its ends are not a
beginning or an ending of the generated symbol
or when there is an adjacency punctuation mark;
it would not be a rejection in the case of subordinate
sentences—the existence of subordinate elements
would be verified as would the existence of multiple
verbal forms.

6.4 Other cases
There are specific situations in which ambiguities
can be resolved starting from considerations of
words, grammatical categories and intervening
objects.

7 Experimental Results

We analysed 776 selected sentences covering the
broadest spectrum of casuistry inherent to Spanish
grammar. The reliability measure for the disambi-
guation is given by:

G ¼ ðp � 100Þ=ðn � 1Þ

where p is the total number of functional combina-
tions minus the number of functional combinations

Functional Disambiguation Based on Syntactic Structures

Literary and Linguistic Computing, Vol. 21, No. 2, 2006 195



accepted and n is the total number of functional
combinations provided by the morphological
analyser.

As can be seen, in Fig. 1 and Fig. 2, the yield
of the functional disambiguation—local and
structural—increases with the number of symbols
of a sentence. The functional disambiguation based
on local syntactic structures has an average yield of
87% and increases to a high of 96% after applying
the structural conditions.

8 Conclusions

This study does not stop in subsets of grammar
but challenges a whole system of rules for Spanish
grammar, despite the notable amount of combina-
tions needed for the analysis.

It contributes towards a solution to the problem
of the emergence of functional ambiguities. First,
a process of disambiguation based on local syntactic
structures is applied; it reaches an average yield of

0
10
20
30
40
50
60
70
80
90

100

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
Sentence symbols number

A
v
e

ra
g

e
 g

o
o

d
n

e
s
s

after local functional disambiguation

after structural functional disambiguation

Fig. 1 Functional disambiguation goodness.

0
5

10
15
20
25
30
35
40
45
50
55
60
65
70
75

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17

Sentence symbols number

A
c
e
p

te
d

 f
u

n
c
ti

o
n

a
l

c
o

m
b

in
a
ti

o
n

s

after local functional disambiguation

after structural functional disambiguation

Fig. 2 Number of accepted functional combinations.

O. S. Suárez et al.

196 Literary and Linguistic Computing, Vol. 21, No. 2, 2006



87%. Subsequently, a disambiguation based on
syntactic representation trees is applied. It improves
the yield or performance up to a high end of 96%.

The importance of this work lies in its signifi-
cant contribution to the development of future
applications:

(1) It accelerates the process of syntactic analysis
by trimming incorrect structures.

(2) It improves the precision of results in the
advanced searching of words.

(3) It allows discarding options not valid in the
information extraction process.

(4) It detects grammatical errors in written
constructs, etc.

References
Bosque, I., Demonte, V., and Lázaro Carreter, F. (1999).

Gramática descriptiva de la lengua española. Madrid:

Espasa.

Gili Gaya, S. (1998). Curso Superior de Sintaxis Española.
Barcelona: Biblograf S.A.

Gómez Torrego, L. (2002). Análisis sintáctico. Teorı́a y
práctica. S.M., Madrid.

Quesada, J. F. (1996). Un modelo robusto y eficiente para
el análisis sintáctico de lenguajes naturales mediante
árboles múltiples virtuales. Centro Informático
Cientı́fico de Andalucı́a (CICA).

Real Academia Española (1989). Esbozo de una nueva
gramática de la lengua española. Madrid: Espasa-Calpe.

Santana, O., Pérez, J., Carreras, F., Duque, J.,
Hernández, Z., and Rodrı́guez, G. (1999). FLANOM:
Flexionador y lematizador automático de formas
nominales. Lingüı́stica Española Actual XXI, 2: 253–97.

Santana, O., Pérez, J., Hernández, Z., Carreras, F., and
Rodrı́guez, G. (1997). FLAVER: Flexionador y
lematizador automático de formas verbales.
Lingüı́stica Española Actual XIX, 2: 229–82.

Santana, O., Pérez, J., Losada, L., and Carreras, F.
(2002). Hacia la desambiguación funcional automática
en Español. Procesamiento del Lenguaje Natural,
28(SEPLN): 1–22.

Functional Disambiguation Based on Syntactic Structures

Literary and Linguistic Computing, Vol. 21, No. 2, 2006 197