Final Spanish Lexicon Documentation
gilcUB - FBG
Universitat de Barcelona
April 1998
Authors:
Nuria Bel, Isabel Brosa, Montserrat Marimon, Marta Villegas
This report describes the current status of the Parole Spanish Lexicon. This lexicon contains about 20,000 entries. These entries had to be selected according to the Corpus also treated in the Parole project. Spanish situation was in this case different, as no corpus for Spanish was finally included. However, trying to follow the general policy of the project, which is to encode general vocabulary coming from real corpora, we have selected the entries according to frequency tables of in-house available corpus of about one million words. In order to reach the final figure we had to add some more lexical entries. The strategy for such a selection was mainly derivational patterns, i.e. if a verb was present in the frequency list, the corresponding nominal and adjectival derivation were also included if missed.
This selection procedure was time consuming, although it was compensated by the fact that conversion mechanisms for transferring in-house morphological resources into SGML-Parole format were rather successful. This success was based in two main facts. First, the design of the Parole model, concerning the Spanish morphology, could maintain the same philosophy as our internal resources. Verbal morphology was almost automatically converted. As for nominal inflection, some hand encoding was needed, but special routines were created for that purpose.
Our internal morphological dictionaries follows the "Word and Paradigm" strategy, then all the stems are stated for a verb. As the Parole model accepted the possibility of having different strategies for morphology generation, we could just convert the format from attribute/value formalism to Parole without problems. For Nominal morphology the conversion was more complex as our internal resources applied the "two level morphology" strategy for converting a base form in different allomorphs. Besides, the morphological classification used had to suffer some changes due to the lack of generalisation coming from the change of the underlying strategy. Once the model was fully developed, we implemented algorithms for re-using the available information (gender, number, type,...), and we add some heuristics to derive the information needed for Parole (i.e. "all nouns ending with "-ción", have the same MFGN and have two allomorphs") Thus, we could automatically produce the two stems of a given class of words automatically. This conversion routines used shell and Perl scripts and performed a 80% of new encoding automatically. The rest was done by hand.
As for syntax, the situation was rather different as no previous resources could be used. However, we used as basis for developing the Parole model for syntax, the work done by our group in other projects such as EUROTRA, TRADE and LS-GRAM. Once all the model and for all the categories was developed, we could start the actual encoding of syntax. It has to be said that the model is complex and it is difficult to get familiarised with it. This justifies the change of internal milestones proposed by our group. We have also to mention that we started encoding verbs, which although they result long to encode, have better documentation than other categories.
As a final remark, we would like to comment on the fact that we have decided to work mainly using SGML format for all the model development and encoding work. Standard filler is thus used for feeding the AlethGD database. Filling exercise has proven to be easy and successful.
We mainly follow the Parole philosophy for modelling Spanish morphology. Hence we will only reproduce here the figures of our lexicon.
| Number of simple morphological units | 22054 |
| Number of graphical morphological units | 22058 |
| Number of simple inflection modes | 132 |
| Number of simple compound inflection modes | 2 |
| Category | Subcategory | Number of Units |
| NOUN | COMMON | 12209 |
| VERB | MAIN | 3367 |
| ADJECTIVE | QUALI | 4996 |
| PRONOUN | DEMONSTRATIVE | 3 |
| PRONOUN | INDEFINITE | 16 |
| PRONOUN | SSCSTRONG | 6 |
| PRONOUN | SSCWEAK | 6 |
| PRONOUN | INTERROGATIVE | 4 |
| PRONOUN | RELATIVE | 4 |
| ADVERB | PARTICLE | 1 |
| ADVERB | GENERAL | 601 |
| ADPOSITION | PREPOSITION | 19 |
| CONJUNCTION | COORDINATIVE | 2 |
| CONJUNCTION | SUBORDINATIVE | 1 |
| NUMERAL | CARDINAL | 3 |
| NUMERAL | ORDINAL | 3 |
| DETERMINER | POSSESSIVE | 5 |
| DETERMINER | DEMONSTRATIVE | 3 |
| DETERMINER | INDEFINITE | 18 |
| DETERMINER | INTERROGATIVE | 3 |
| ARTICLE | INDEFINITE | 1 |
| ARTICLE | DEFINITE | 1 |
In the following pages we will present the syntactic constructions we have developed as well as the criteria applied to model syntactic information. These criteria are grouped along categorial dimension.
Any morphological unit has assigned at least one syntactic unit (SynU) which bears syntactic information (essentially, concerning subcategorization information). Many often a lexical item has more than one subcategorization frame (i.e. syntactic description). In these cases, we have to decide how to organise this valence alternation. PAROLE provides us with different strategies or mechanisms to express frame variation within a given lexical entry. These are:
1)- Grouping alternative phrases within a given position within the same construction. In this case, it is required that both phrases have the same syntactic function (and, possibly the same thematic role) and that they undergo the same syntactic processes. So, for instance, the Spanish object position can be filled in by a full NP, a [+animated] NP introduced by the preposition ‘a’ and a weak pronoun in accusative. This alternation is encoded at the level of position.
2)- Encoding different descriptions (a base description and one (or more) derived description) within a unique SynU (Syntactic Unit). These descriptions may be related or not by means of a frame set.
The approach followed in the Spanish lexicon is a ‘structured’ one. We aim at modelling syntactic information in a way which is sensitive with linguistic relevant phenomena. Our syntactic information is, therefore, structured.
Essentially we follow the principle: for each subcategorization frame assign one SynU provided the frame alternation involved is particular to that lexical item. Frame alternations which are not idiosyncratic but common to a well defined set of verbs are grouped together in the same SynU and related by means of a relevant FrameSet. Valence alternations encoded via FrameSet include:
(i) Passivization: active / passive descriptions are included in the same SynU and related by a (passive) FrameSet
(ii) Pronominalization derived from a decausative process. Transitive causative and intransitive pronominalized (decausative) descriptions are included in the same SynU and related by a FrameSet as exemplified above.
This excludes ‘real’ reflexive pronouns and ‘aspectual pronominalization’ exemplified below:
‘Aspectual pronominal’ verbs have an additional SynU and the pronominal description is never related to the non-pronominalized description.
(iii) Decausativization: some Spanish verbs can decausativize without resorting to ‘se’ pronominalization:
In this case the transitive / intransitive descriptions are included in the same SynU and related by an (ergative) FrameSet.
(iv) Equi relations: Most Spanish control verbs have a valence alternant with a finite complement. This alternation is also encoded by means of a (equi) FrameSet.
(v) Symmetric predicates: frame alternation involving symmetric predicates (subject and object) is also encoded via FrameSet:
(vi) Optionality: except for subject ‘pro-drop’ and optionality of By-agents in passive constructions, optionality is encoded by means of a (deletion) FrameSet. So, ‘valence reduced’ descriptions are related to their ‘non-valence reduced’ counterparts by means of a (deletion) FrameSet. As far as direct object position is concerned, object deletion strictly corresponds to transitive/absolutive readings (eg: comer patatas vs comer a las dos). Optionality involving IOs and prepositional complements is much more flexible.
Subject Pro-drop is not explicitly encoded, thus subject position is not marked as OPTIONAL=YESO. Optionality of By-agents is encoded at the level of position.
FrameSets do not deal with:
Summarising up, the resolution algorithm followed in the Spanish lexicon goes as follows:
In verbal descriptions the verb is always assigned the HEAD function in a construction (clause) with zero (for impersonal verbs) or more positions according to the verb’s subcategorization requirements. Verbal descriptions, therefore, describe verb’s maximal projections.
Self element in verbal constructions:
As far as verbal descriptions are concerned, we have distinguished up to six kinds of self objects. These include: active, passive, pronominal, copulative, auxiliary, and ‘terciopersonal’:
Auxiliary verbs only include ‘ser’ and ‘estar’. Other ‘auxiliary-like’ verbs such as ‘seguir’, ‘continuar’ and modal verbs are not encoded as MORPHSUBCAT=AUX.
Pronominal verbs are those which require a reflexive which can not be taken as part of the argument structure of the verb. This includes ‘pure pronominal’ verbs such as ‘acordarse’, decausative verbs such as ‘abrirse’, ‘aspectually-marked’ verbs such as ‘comerse, irse, marcharse...’.
In the following lines we give the criteria we have followed when encoding verbal complements:
Finite complements:In Spanish, sentential complements may 'occur' as subjects, objects or as oblique complements introduced by a preposition:
In all cases, the sentential complement is a Syntagma labelled as 'Clause' filling in the SUBJECT, OBJECT and OBLIQUE position respectively. Oblique sentential complements (that is, those introduced by strongly bound prepositions) are taken as 'Clauses' where the preposition functions as a 'marker'. Strongly bound prepositions are encoded by means of the LexicalFeature attribute.
We start distinguishing sentential complements which alternate with infinitive complements in control verbs from other sentential complements. Sentential complements in control verbs are always introduced by 'que'. This is encoded by means of the attribute/value SYNSUBCAT=THATCL.
In addition, sentential complements of control verbs are further specified according to their +/-control +/-obviative nature and their mood. Controlled and obviative sentential complements bear the attribute/value CORREF=COJ and COREF=NOCOJ respectively. The mood of sentential complements is encoded by means of the MOOD attribute.
The set of sentential complements in control verbs includes:
VP/SCOMP alternation in control verbs is encoded by means of 'Equi Framesets'. The infinitive frame is always the 'base description' whereas the finite frame is the 'derived description':
VP/SCOMP alternation in non-strictly control verbs such as 'confesar' is not encoded via Frameset. These kinds of verbs admit indicative and subjunctive sentential complements which may be introduced by different elements:
Unfortunately, we have not been able to define the set of sentential complements in non-control verbs. This led us to take a 'save approach'. We define the basic 'Clause' syntagma above which, being underspecified and less informative, is taken a supertype which subsumes 'that clauses' of control verbs (more informative) and sentential complements of ‘non-control’ verbs.
The Strategy to describe control information
In control constructions, the control verb is the self/head element of a construction with two (or more positions), one corresponding to the controller element an the other to the verbal infinitive complement. The VP complement is defined as having an OPENLIST with a subject position valued as Coref=COI. This value is ‘shared’ by the controller element:
In obviative constructions the ‘embedded’ subject is valued Coref=NOCOI, this indicates that the embedded subject and the matrix element have distinct Coref value.
Indirect Object vs Direct Objects and 'Benefactives':
All Spanish verbs admit a complement which cliticizes in dative. Not all dative clitizable complements are 'true IOs', often these complements are 'benefactive' or 'ethic' complements. Only 'real' IOs are encoded in the Spanish lexica. We assume the grammar generalises the fact that all verbs can be attached a benefactive.
Often, benefactives are evident:
Sometimes, however, benefactives are not so evident and may be taken as IOs. This is the case of what we call 'mental or body part' complements:
These complements are not considered IOs but 'benefactives'. All ‘benefactive’ complements can be substituted by a PPde complement (in fact, ‘benefactive’ complements and PPde complements are in complementary distribution):
Many often 'benefactive' complements and DOs require the same referent. Note the ungrammatical examples below where the two complements have a distant referent:
Direct Object vs Indirect Object:
In Spanish clitic case has been taken as a criteria to distinguish DOs and IOs. Thus, DOs cliticize in accusative whereas IOs cliticize in dative.
Many often, however, it is not clear whether a given complement is a DO or an IO. The fact that not all transitive verbs allow passivization together with the fact that certain accusative and dative clitic forms coincide may lead to confusion (also 'leismo' and ‘laismo’ complicates things). This is the case of bivalent verbs with a +human complement. In the examples below we have the same structure:
In the following lines we list the criteria we have followed when distinguishing between DOs and IOs:
For other verbs, passive construction improves:
In some way, this PPcon contributes to the agentiveness of the subject and may be understood as 'body part' or 'mental part':
and as complements of object control adjectives:
Note that the reading for 'se-pronominalized' 'IO verbs' is the reflexive reading, never the decausative reading. In fact 'IO verbs' are in some way decuasative verbs. To express a causative reading we have to use other verbs:
whereas 'IO verbs' hardly admit benefactive complements.
In the following lines we list the set of verbal descriptions defined in the Spanish group. They are grouped according to the nature of the Self element (that is, according to whether the verb is ‘active’, ‘passive’, ‘pronominal’ or ‘auxiliary’):
Description_id: description: example:
Descriptions where the Self occurs in active form:
Descriptions where the Self occurs in passive form (we obviate the optional by-agent position):
Descriptions where the Self is pronominalized (pronominal verbs are those where the reflexive pronoun is not part of the argument structure of the verb but rather it is taken as part of the verb, they include 'pure pronominal’ verbs such as 'acordarse', pronominal decausative verbs such as ‘abrirse’, and 'aspectual' pronominal verbs such as 'comerse'):
Copulative verbal descriptions:
Impersonal and ‘terciopersonal’ descriptions:
The modelling of syntactic information in adjectives is similar to that of verbs. Thus, preposition alternation and category alternation involve splitting of descriptions into different SynUs: This is the case of the adjective 'atento a' (attentive to) vs 'atento con' (considerate with):
or the adjective ‘seguro’ (sure) which can takes nominal and infinitival (control) complements:
Unlike verbs, however, optionality of prepositional complements is not encoded via deletion FrameSets relating ‘complete’ and ‘valence reduced’ descriptions, but rather it is encoded at the level of position. Thus, adjectival prepositional complements are regarded as ‘marked’ NPs filling in optional positions.
Non-predicative adjectives can not occur as complement to the copula. In Spanish, all non-predicative adjectives occur in prenominal position (ex: ‘el mero hecho de su llegada’). Some of them can also occur in postnominal position. This information is encoded in the SELF by means of a SyntFeatureOpen feature we name POSADJ whose values are PREN (for prenominal) POSTN (for postnominal) and PREPOSTN (for those adjectives which can occur in both positions). Non-predicative adjectives have Empty constructions (that is, constructions with zero positions).
Predicative adjectives can occur as complement to the copula. In Spanish we have to distinguish between 'ser' adjectives and 'estar' adjectives according to whether they occur with ‘ser’ or ‘estar’ copula verbs. This information is encoded in the SELF by means of an SyntFeatureOpen we name COPU whose values are SER or ESTAR. Some adjectives can occur in both 'ser' and 'estar' constructions. Since such alternation generally implies a difference in meaning, we give two different descriptions.
In these cases, we have two SynUs one for the 'ser reading' and another for the 'estar reading'.
The SELF element in predicative adjectives distinguishes between 'estar' vs. 'ser' and 'pre-nominal' vs. 'post-nominal'. All predicative adjectives are the HEAD/SELF element of a construction with a SUBJECT position (i.e. external argument) and valued for COREF. This obligatory SUBJECT position is occupied by a controlled subject (the same controlled subject as in the case of control verbs) and takes the same COREF value of the construction:
| Predicative ‘ser’ Adjective | ||
| Ex. : un poder ilimitado [SYNTLABEL:AP, CORE:COJ] | ||
| - |
SELF |
POSITION |
| Syntlabel | A | NP |
| Posadj | POSTN | -- |
| Copu | SER | -- |
| Function | HEAD | SUBJECT |
| Coref | COJ | COJ |
The COREF feature in the construction encodes its 'predicative' nature. So, copula verbs are defined as the self element in a construction with two positions, a subject position and a subject-predicative position, which are COREF valued:
| Copula verb: | |||
| Ex. Juan es alto / capaz de no venir [SYNTLABEL: Clause] | |||
| - |
POSITON |
SELF |
POSITION |
| Syntlabel | NP | V | AP |
| Function | SUBJECT | HEAD | SUBJPRED |
| Morphsub. | -- | COPULA | -- |
| Copu | -- | SER | SER |
| Coref | COJ | -- | COJ |
The use of the COREF feature in the SUBJPRED position above avoids having to (further) describe the Adjectival Phrase looking for its controlled SUBJECT (or OBJECT, in case of object control adjectives) and allows us to approach all control constructions (verbal and adjectival) in the very same terms.
Prototypically, subjects of predicative adjectives are unexpressed. In certain constructions, however, the subject may be lexically realised as a NP introduced by the preposition ‘de’:
In this case, the SUBJECT position is occupied by a NPde.
Most predicative adjectives allow for an optional specifier. In the case of adjectives, specifiers are of typically degree nature. Adjectives admitting for specifiers have and optional position (function=AMODIFIER) occupied by a 'degree' adverbial phrase:
Some predicative adjectives take a prepositional complement. The preposition introduces a NP complement or a VP complement. In the former case, the construction has an optional position (function=APREPCOMP) filled by a 'marked' NP (remember we do not take strongly bound prepositional complements as PPs but rather as marked NPs or VPs): (we obviate here the adjectival modifier position)
Predicative adjectives taking infinitive complements express an obligatory control relation. Control adjectives can be subject control or object control:
For control adjectives, the APREPCOMP position is filled in by a subject controlled VP (as in the case of control verbs):
| ‘Control Adjectives’: | ||||
| Ex. ‘fácil, seguro’ | ||||
| -- |
SELF |
POSITION |
POSITION |
POSITION |
| Syntlabel | ADV | ADV | NP | VP |
| Synsubcat | -- | DEGREE | -- | -- |
| Function | HEAD | AMODIFIER | SUBJECT | APREPCOMP |
| Copu | SER v ESTAR | -- | -- | -- |
| Posadj | POSTN | -- | -- | -- |
| Coref | -- | -- | COJ | COJ |
| Mood | -- | -- | -- | INFINITVE |
| LexFeature | -- | -- | -- | Prep: DE |
| Optional | -- | YESO | NOO | NOO |
Subject control adjective differ from object control adjectives in the internal description of the embedded VP in position2. In both cases, the APREPCOMP position is filled by a 'controlled' VP[de] defined as OPENLIST=YES. In the first case the subject position is defined as COREF=COJ whereas in the latter case the object position is defined as COREF=COJ. In the following tables we give the attributes/values used to define these ‘Syntagmas’ and their ‘Positions’:
| Subject controlled VP: | ||
| - |
SYNTAGMA |
POSITION |
| Syntlabel | VP | NP |
| Function | APREPCOMP | SUBJECT |
| Mood | INFINITIVE | -- |
| LexFeature | Prep:DE | -- |
| Coref | COJ | COJ |
| ‘Object controlled VP’: | |||
| -- |
SYNTAGMA |
POSITION |
POSITION |
| Syntlabel | VP | NP | NP |
| Function | HEAD | SUBJECT | OBJECT |
| Mood | INFINITIVE | -- | -- |
| LexFeature | Prep: DE | -- | -- |
| Coref | COJ | -- | COJ |
| Morpfhubcat | GENERAL | -- | -- |
Irregular Comparative adjectives:
In Spanish comparative adjectival constructions take the 'más/menos + adj + que' form:
A small number of (irregular) adjectives have comparative synthetic forms:
Irregular comparative adjectives are the HEAD/SELF of a construction with:
(i) an optional AMODIFIER position (comparative adjectives admit a degree specifier as in ‘'mucho mejor que Juan' ‘more better than Juan’). This AMODIFIER position differs from the AMODIFIER position of non-comparative adjectives as it does not allow for the same set of degree adverbs:
This information is encoded in Adverbial descriptions.
(ii) An external subject position (remember that all predicative elements have an ‘external’/subject argument position).
(iii) An ACLAUSCOMP position filled in by a comparative clause introduced by ‘que’.
| ‘Irregular comparative adjectives’: | ||||
| Ex. ‘(mucho) mejor (que tú). | ||||
| -- |
SELF |
POSITION |
POSITION |
POSITION |
| Syntlabel | ADJ | ADV | NP | CL |
| Synsubcat | COMPARATIV. | DEGREE | -- | ThatCL SCOMP |
| Function | HEAD | AMODIFIER | SUBJECT | ACLCOMP |
| Optional | -- | YESO | NOO | YESO |
| Copu | SER | -- | -- | -- |
In the following lines we list all descriptions for the category ADJECTIVE (we give the Parole_id, an example, the number of occurrences and the corresponding construction):
For all nominal descriptions, the noun is the head/self of a construction with zero, one, two or three positions depending on the type of the noun. As far as the head/self is concerned, nominal descriptions are classified into 'proper nouns' nouns and 'common nouns'. They are distinguished by means of the 'featurename' MORPHSUBCAT. Common nouns are further classified into Countable, Uncountable and Mass by means of the OpenFeature NTYPE valued as COUNTABLE UNCOUNTABLE or MASS. See below for the criteria used. Nominal constructions describe the set of positions demanded by nouns. These positions can be filled by determiners as well as by nominal arguments.
Proper Nouns: The encoding of determiners:
Proper nouns are classified according to whether they admit a definite article or not. Proper nouns with no article are defined as the head/self of an 'empty' (no positions) construction:
Proper nouns requiring a definite article are defined as the head/self of a construction with a position0 filled in by an obligatory determinative phrase specified as MORPHSUBCAT=DEFINITE and MORPHCAT=ARTICLE:
| Proper Nouns with definite article | ||
| Ex. : ‘Los Angeles’ | ||
| -- |
SELF |
POSITION |
| Syntlabel | N | ARTICLE |
| Morphsubcat | PROPER | DEFINITE |
| Function | HEAD | NDETERMINATIVE |
| Optional | -- | NOO |
All common nouns have an optional position0/NDETERMINATIVE filled by a determinative phrase (DetP). The kind of DetP a given noun admits depends on the type of the noun.
As we have already mentioned before we classify non-predicative common nouns as COUNT, UNCOUNT or MASS. This classification tries to capture the different distributional behaviour of nouns. Nouns differ as far as the environment they can occur in. Thus, for instance, certain nouns can not be preceded by numerals whereas others can. Not all distributional differences are explicitated in the Spanish PAROLE lexicon. However, the classification adopted can be used to generalise their distributional behaviour.
Nouns are classified into three types depending on the kind of DetP they co-occur with. The variety and types of DetPs is very huge. Nouns can be preceded by a great variety of premodifier constructions. Here we follow a 'prototype based' approach. Thus, we are only concerned about +/-countable nature of nominal denotation (premodifier constructions are fully described in determiners).
The criteria listed below refers to a prototypical usage. Nouns may 'change' their type when used metaphorically or even in certain sublanguages. ‘Recategorization’, therefore, is not encoded.
(i) have plural/singular distinction:
(ii) when used in singular, they need a determiner (compare with [-countable] nouns):
(iii) only plural forms can normally occur in object position with no-determiner:
(iv) can be 'enumerated'. Can co-occur with numerals:
(v) when used with quantifiers such as 'poco, bastante, or nada de' need to occur in plural:
(vi) similarly do not co-occur with 'partitive' elements:
(i) cannot be enumerated:
(ii) cannot be quantified by 'distributional' quantifiers such as ‘cada, cualquier, ambos, algun(), ningun, cierto(), mismo()’.
(iii) cannot be 'fractionated' :
(i) cannot be enumerated:
Some mass nouns admit 'recategorization' and admit a 'countable' usage:
Note, however, that this is a 'countable' usage of an prototypically '-countable' noun. This usage is not explicitly encoded, but can be easily derived.
(ii) when quantified, they are used in singular:
(ii) they tend to avoid plural forms:
(iii) article usage: Unlike countable nouns, mass nouns cannot occur with article in constructions 'esto es...':
(iv) can occur with partitive specifiers
They can be used as countable and mass nouns in that:
a) can be used in singular and plural after a determiner (object position):
b) can occur with and without article in 'esto es...' constructions:
c) when quantified, they can be used in singular and in plural:
d) can be used with 'distributional' quantifiers and with partitives:
Variable nouns are assigned one SynU with two descriptions (a 'Countable' description and a 'Mass' description ).
Predicative nouns are not classified according to their +/- countable nature, therefore, they bear an 'under-specified' DetP (the DetP is not further labelled). Only non-predicative common nouns are further classified according to the kind of DET they admit:
| Countable Noun | ||
| Ex. : ‘niño, libro, ' | ||
| -- |
SELF |
POSITION |
| Syntlabel | N | DET NUM ART |
| Morphsubcat | COMMON | DEM POSS INDEFdistrib |
| Function | HEAD | NDETERMINATIVE |
| Ntype | COUNTABLE | -- |
| Optional | -- | YESO |
| Uncountable Noun | ||
| Ex. : ‘paz, fe, agricultura, esperanto' | ||
| -- |
SELF |
POSITION |
| Syntlabel | N | DET ART |
| Morphsubcat | COMMON | DEM POSS |
| Function | HEAD | NDETERMINATIVE |
| Ntype | UNCOUNTABLE | -- |
| Optional | -- | YESO |
| Mass Noun | ||
| Ex. : ‘leche, agua, petróleo' | ||
| -- |
SELF |
POSITION |
| Syntlabel | N | DET ART |
| Morphsubcat | COMMON | DEM POSS INDEFglobal |
| Function | HEAD | NDETERMINATIVE |
| Ntype | MASS | -- |
| Optional | -- | YESO |
| Plural Noun | ||
| Ex. : ‘víveres, catacumbas' | ||
| -- |
SELF |
POSITION |
| Syntlabel | N | DETP |
| Morphsubcat | COMMON | -- |
| Function | HEAD | NDETERMINATIVE |
| Optional | -- | YESO |
| Tnumber | PLURAL | PLURAL |
The encoding of nominal arguments:
Appositions: A number of common nouns admit an apposition (eg ‘el señor González’). In this case the noun bears an additional position1/NAPPOSITION filled by a NP.
Predicative nouns: Most Spanish nouns require for an optional prepositional phrase introduced by 'de'. Although these constructions are structurally identical (that is, N+PPde), semantically, they may be different as the deep syntactic function (and the assigned thematic role) of this PP[de] element differs from noun to noun. Thus, this PP[de] may 'correspond' to
a) the verb's subject of deverbal nouns:
b) the verb's object of deverbal nouns:
c) the 'promoted' subject in decausative readings:
d) collective and quantifiers:
e) the adjective's subject in deadjectival nouns:
f) Sometimes this PP[de] may correspond to different semantic arguments...
g) in all cases the PPde is in complementary distribution with a possessive pronoun:
Despite the different semantic nature of the prepositional element in the N+PPde constructions above, the fact is that in all cases we have the same structural construction. This allows us to define a single description where the noun is the self/HEAD of a construction with an optional position0/NDETERMINATIVE filled by a DetP and an additional position1/NCOMP filled by the NP[de] complement.
Despite NP[de] complements are always optional, this optionality is NOT encoded in the position (i.e. by means of the feature OPTIONAL=YES). Instead, we duplicate SynUs so that there is one SynU with no complement NP[de] and another with NP[de] complement. (see motivations below)
Not all PPde are taken as complements, some are considered modifiers and, therefore, are not encoded:
The rest of nominal prepositional complements differ from PPde complements in that they are assigned NPREPCOM function. This distinction is fully motivated as Romance PPde nominal complements have a special status.
Verbal complements: Certain nouns may take, besides the optional PPde complement, a verbal complement introduced by a preposition. The status of this verbal complement varies according to its (non)controlled nature.
Support verbs: Some nouns can occur in ‘support verb’ constructions:
When predicative nouns occur in support verb constructions, the NP[de] is not allowed as it 'occurs' as subject of the verb (compare the examples above). Such complementary distribution forbids encoding NPde optionality at the level of position. Support verb information is included in the example attribute.
In the following lines we list all descriptions for the category NOUN (the table includes the Parole_id, an example, the number of occurrences and the construction):
Attribute/values used when describing the SELF object in Adverbials:
SYNSUBCAT: DEGREE for 'degree' adverbs such as 'muy' and 'bastante'.
FUNCTION: Prototypically adverbs are modifiers. Adverbs are modifiers in constructions where the HEAD may be a Noun, a Verb, an Adjective or another Adverb. This means that, in adverbial descriptions, the SELF object is not assigned HEAD function but rather XMODIFIER function.
Unlike verbs, nouns or adjectives, the 'insertion context' for adverbs goes beyond its 'maximal projection': the SYNTLABEL value of a construction whose SELF object is an adverb is a NP, AP, ADVP, VP or Clause depending on whether the assigned HEAD is a Noun, Adjective, Adverb, Verb or Clause respectively.
As far as construction/environment is concerned, Adverbs are classified/described according to four aspects:
1. the element they occur with and modify. (that is, Noun, Adjective, Verb, Adverb or Sentence):
Vmodifier and Smodifier adjectives can be easily distinguished: Verbal adjuncts not only occur in topicalizations and wh-questions but also in all other constructions involving unbounded dependencies such as indirect questions, relatives and clefts:
Sentential adverbs and Verbal adverbs can co-occur (something impossible if both adverbials are Vmodifier and belong to the same semantic class):
2. whether they admit a modifier:
Adverbs which admit an optional adverbial modifier bear an optional position0/ADVMODIFIER filled in by a ADVP specified as ‘degree’.
3. whether they require a complement (only for deadjectival adverbs ending in '-mente', see below):
Unlike other adverbs, 'deadjectival' adverbs with a PP complement are heads of a ADVP. (note that other adverbs are modifiers of constructions where the head may be a V Adj or Adv depending on the type of adverb).
4. whether they admit an apposition:
Comparative adverbs are the self/AMODIFIER element in a AP construction with a position0/HEAD filled by an ADJ and an optional position1/ACLAUSCOMP filled in by a comparative clause.
| ‘Comparative’: | |||
| Ex. ‘más’ AP: [AMODIFIER más / menos] [HEAD alto] (ACLAUSCOMP que Juan) | |||
| -- |
SELF |
POSITION |
POSITION |
| Syntlabel | ADV | AP | Clause |
| Synsubcat | DEGREE | -- | ThatCL SSCOMP |
| Function | AMODIFIER | HEAD | ACOMP |
| Whtype | WHNO | -- | -- |
| Ndegree | POSITIVE | -- | -- |
| Morpfhubcat | GENERAL | -- | -- |
‘Intensifier’ adverbs are the self/ADVMODIFIER element in a AP construction with a position0/AMODIFIER filled in by an ADV, a position1/HEAD filled in by an ADJ and an optional position2/ACLAUSCOMP filled in by a comparative clause (we only give relevant information):
| ‘Intensifier adverb’: | ||||
| Ex. ‘mucho, bastante’ | ||||
| -- |
SELF |
POSITION |
POSITION |
POSITION |
| Syntlabel | ADV | ADV | AP | CL |
| Synsubcat | DEGREE | DEGREE | -- | ThatCL SCOMP |
| Function | ADVMODIFIER | AMODIFIER | HEAD | ACOMP |
‘Intensifier’ adverbs can also occur as modifiers of ‘inflected’ comparative adjectives. In this case they are the self/AMODIFIER element in a AP constructions with a position0/HEAD filled in by a ‘inflected’ comparative ADJ and an optional position1/ACLAUSCOMP filled in by a comparative clause:
| ‘Intensifier comp. Adverbs’: | |||
| Ex. ‘mucho, bastante’ | |||
| -- |
SELF |
POSITION |
POSITION |
| Sy | |||