Electronic Resources for NLP LexiconGrammar and Semantic Predicates

What is Lexicon-Grammar? With Lexicon-Grammar we mean the method and the practice of formal

Rules and Exceptions In order to define what can be considered to be a

“Grammar, or, as it has now been called, linguistic theory, has always been driven

“But grammarians operating at the level of sentences seem to be interested only in

LG Methodology In the LG methodology it is crucial the collection and the analysis

The necessity of local grammars What emerges from the LG studies is that, associating

LG Classification The LG classification and description of the Italian verbs is grounded on

LG Classification Every LG class has its own definitional structure, that corresponds with the

LG Classification All the lexical entries are, then, differentiated one another in each class,

Distributional Analysis The distribution of an element A is defined by Harris as the

Distributional Analysis The selection depends on the probability that lexical elements, that belong to

Distributional Analysis When two sequences can appear in the same context they have an

Trasformational Analysis The sentences A and B, despite a different combination of words, are

Trasformational Analysis A transformation is defined as a function T that ties a set

Trasformational Analysis Transformational relations must not be mixed up with any derivation process of

Trasformational Analysis ◉ Substitution ◉ Paraphrases with support verbs, nominalizations and adjectivalizations. nominalization (e.

Lexicon Grammar Tables distributional properties, bywhich all the co-occurrences of the lexical items L

Lexicon Grammar Tables http: //dsc. unisa. it/composti/tavole/combo/tavole. asp

The Sintax-Semantics Interface In the sixties, the hot debate about the syntax-semantics Interface opposed

The Sintax-Semantics Interface Chomsky (1957), postulated the independence of the syntax from the semantics.

The Sintax-Semantics Interface The Lexicon-Grammar framework offers the opportunity to create matches between sets

“Soit Sy l’ensemble des formes syntaxiques d’une langue [. . . ]. Soit Se

The Semantic Predicate Theory Gross, M. (1981). Les bases empiriques de la notion de

The Semantic Predicate Theory Correspondence between some element from Sy and a semantic predicate

The Semantic Predicate Theory Gross’s 1981 purpose was to demonstrate the uselessness of the

The Semantic Predicate Theory ◉ The parallelism between the Sy and the Se that,

The Semantic Predicate Theory ◉ Lexical items are not a priori predicative or argumental:

The Sintax-Semantics Interface The richness of the syntactic-distributional properties allows the exploration of semantic

The Semantic Predicate Theory The possibility to create intuitive semantic macro-classifications into the LG

The Sintax-Semantics Interface The syntactic transformations in which the same predicate is involved preserve

The Sintax-Semantics Interface Il sistema LEG-Semantic Role Labeling (LEG-SRL) per l’italiano Elia, A. (2014

Slides: 50

Download presentation

Electronic Resources for NLP: Lexicon-Grammar and Semantic Predicates

Lexicon-Grammar

What is Lexicon-Grammar? With Lexicon-Grammar we mean the method and the practice of formal description of the natural language, introduced by Maurice Gross in the second half of the 1960 s, who, during the verification of some procedures from the transformational-generative grammar (TGG) (Chomsky, 1965) laid the foundations for a brand new theoretical framework. A complex idea can be conveyed with just a single still image, namely making it possible to absorb large amounts of data quickly.

Rules and Exceptions In order to define what can be considered to be a rule and what must be considered an exception, a small sample of the lexicon of the analyzed language, collected into arbitrary ways, can be truly misleading. It is assumed to be impossible to make generalizations without verifying or falsifying the hypotheses.

“Grammar, or, as it has now been called, linguistic theory, has always been driven by a quest for complete generalizations (…). This development contrasts with that of the other Natural Sciences such as Biology or Geology, where the main stream of activity was and is still now the search and accumulation of exhaustive data” “ Gross, M. (1997). The construction of local grammars. In Finite-state language processing, volume 11, page 329. MIT Press.

“But grammarians operating at the level of sentences seem to be interested only in elaborating general rules and do so without performing any sort of systematic observation and without a methodical accumulation of sentence forms to be used by further generations of scientists” “ Gross, M. (1997). The construction of local grammars. In Finite-state language processing, volume 11, page 329. MIT Press.

LG Methodology In the LG methodology it is crucial the collection and the analysis of a large quantity of linguistic facts and their continuous comparison with the reality of the linguistic usages, by examples and counter-examples. The collection of the linguistic information is constantly registered into LG tables that cross-check the lexicon and the syntax of any given language. The ultimate purpose of this procedure is the definition of wideranging classes of lexical items associated to transformational, distributional and structural rules.

The necessity of local grammars What emerges from the LG studies is that, associating more than five or six properties to a lexical entries, each one of such entries shows an individual behavior that distinguishes it from any other lexical item. it is always possible to organize a classification around at list one definitional property, that is simultaneously accepted by all the item belonging to a same LG class and, for this reason, is promoted as distinctive feature of the class.

LG Classification The LG classification and description of the Italian verbs is grounded on the differentiation of three different macroclasses: ◉ Transitive Verbs ◉ Intransitive Verbs ◉ Complement Clause Verbs

LG Classification Every LG class has its own definitional structure, that corresponds with the syntactic structure of the nuclear sentence selected by a given number of verbs ◉ V for piovere “to rain" and all the verbs of the class 1 ◉ N 0 V for bruciare “to burn" and the other verbs of the class 3 ◉ N 0 V da N 1 for provenire “to come from" and the verbs belonging to the class 6; etc.

LG Classification All the lexical entries are, then, differentiated one another in each class, by taking into account all the transformational, distributional and structural properties accepted or rejected by every item.

Distributional Analysis The distribution of an element A is defined by Harris as the sum of the contexts of A. Where the context of A is the actual disposition of its cooccurring elements. It consists in the systematic substitution of lexical items with the aim of verifying the semantic or the transformational reactions of the sentences.

Distributional Analysis The selection depends on the probability that lexical elements, that belong to specific classes, can occur together in the same linguistic context. All that is governed by verisimilitude rules that involve a graduation in the acceptability of the item combination.

Distributional Analysis When two sequences can appear in the same context they have an equivalent distribution When they do not share any context they have a complementary distribution.

Trasformational Analysis The sentences A and B, despite a different combination of words, are equivalent from a semantic point of view e. g. Floria ha letto quel romanzo = quel romanzo è stato letto da Floria “Floria read that novel = that novel has been read by Floria”

Trasformational Analysis A transformation is defined as a function T that ties a set of variables representing the sentences that are in paraphrastic equivalence (A and B are partial or full synonym) and that follow the morphemic invariance (A and B posses the same lexical morphemes).

Trasformational Analysis Transformational relations must not be mixed up with any derivation process of the sentence B from A and vice versa. A and B are just two correlated variants inwhich equivalent grammatical relations are realized

Trasformational Analysis ◉ Substitution ◉ Paraphrases with support verbs, nominalizations and adjectivalizations. nominalization (e. g. amare = amore “to love = love”) adjectivalization (e. g. amare = amoroso “to love = loving”) ◉ Repositioning. Emphasis, extraction ◉ Extension ◉ Restructuring. ◉ Passivization ◉ Systematic correlation permutation, dislocation,

Lexicon Grammar Tables

Lexicon Grammar Tables distributional properties, bywhich all the co-occurrences of the lexical items L with distributional classes and subclasses are tested, inside acceptable elementary sentence structures; structural and transformational properties, through which all the combinatorial transformational possibilities, inside elementary sentence structures, are explored (see Section 2. 1. 1 for more in-depth explanations).

Lexicon Grammar Tables http: //dsc. unisa. it/composti/tavole/combo/tavole. asp

Lexicon Grammar Tables

Semantic Predicate Theory

The Sintax-Semantics Interface In the sixties, the hot debate about the syntax-semantics Interface opposed theories supporting ◉ theories supporting the autonomy of the syntax from the semantics ◉ theories sustaining the close link between syntax and semantics

The Sintax-Semantics Interface Chomsky (1957), postulated the independence of the syntax from the semantics. It is due to: ◉ the vagueness of the semantics ◉ the correspondence mismatches among sintax and semantics He recognized the existence of relation among the two layers, that deserves to be deepened

The Sintax-Semantics Interface The Lexicon-Grammar framework offers the opportunity to create matches between sets or subsets of lexico-syntactic structures and their semantic interpretations. Gross, M. (1981). Les bases empiriques de la notion de prédicat sémantique. In Langages, pages 7– 52.

“Soit Sy l’ensemble des formes syntaxiques d’une langue [. . . ]. Soit Se un ensemble d’éléments de sens, les éléments de Sy seront associés à ceux de Se par des règles d’interprétation. [. . . ] nous appellerons actants syntaxiques les sujet et complément(s) du verbe tels qu’ils sont décrits dans Sy; nous appelons arguments les variables des prédicats sémantiques. Dans certains exemples, il y a correspondance biunivoque entre actants et arguments, entre phrase simple et prédicat” “ Gross, M. (1981). Les bases empiriques de la notion de prédicat sémantique. In Langages, pages 7– 52.

The Semantic Predicate Theory Gross, M. (1981). Les bases empiriques de la notion de prédicat sémantique. In Langages, pages 7– 52.

The Semantic Predicate Theory Correspondence between some element from Sy and a semantic predicate from Se ◉ Predicate with three variables: T (p, q, o) ◉ Sentence Structure: «N 0 donne N 1 à N 2» ◉ Interpretation Rules: p = N 0, o = N 1, q = N 2 Gross, M. (1981). Les bases empiriques de la notion de prédicat sémantique. In Langages, pages 7– 52.

The Semantic Predicate Theory Gross, M. (1981). Les bases empiriques de la notion de prédicat sémantique. In Langages, pages 7– 52.

The Semantic Predicate Theory Gross’s 1981 purpose was to demonstrate the uselessness of the semantic representation in predicates and arguments Anyway, it is possible to develope intuitive semantic representation in predicates and arguments able to take into account the granularity of the lexical properties Gross, M. (1981). Les bases empiriques de la notion de prédicat sémantique. In Langages, pages 7– 52.

The Semantic Predicate Theory ◉ The parallelism between the Sy and the Se that, differently from other approaches, is complex ◉ The parallelism do not ignore the idiosyncrasies of the lexicon, since “il n’existe pas deux verbes ayant le même ensemble de propriétés syntaxiques” Gross, M. (1981). Les bases empiriques de la notion de prédicat sémantique. In Langages, pages 7– 52.

The Semantic Predicate Theory ◉ Lexical items are not a priori predicative or argumental: it is the relation among the elements that determines their roles ◉ The minimum semantic units are the elementary sentences, not the words Elia, A. (2014 b). Operatori, argomenti e il sistema leg-semantic role labelling dell’italiano. In Mirto I. , a cura di, Le relazioni irresistibili, pages 105– 118. Edizioni Ets.

The Sintax-Semantics Interface The richness of the syntactic-distributional properties allows the exploration of semantic classifications endowed with a robust syntactic ground The granularity of the verb properties are evidence of ◉ the strong dependence of the syntax from the lexicon ◉ its absolute independence from semantics. Elia, A. (2014 b). Operatori, argomenti e il sistema leg-semantic role labelling dell’italiano. In Mirto I. , a cura di, Le relazioni irresistibili, pages 105– 118. Edizioni Ets.

The Semantic Predicate Theory The possibility to create intuitive semantic macro-classifications into the LG framework does not contrast with the autonomy of the semantics from the syntax Elia, A. (2014 b). Operatori, argomenti e il sistema leg-semantic role labelling dell’italiano. In Mirto I. , a cura di, Le relazioni irresistibili, pages 105– 118. Edizioni Ets.

The Sintax-Semantics Interface The syntactic transformations in which the same predicate is involved preserve the role played by its arguments. Pelosi S. Detecting Subjectivity through Lexicon-Grammar Strategies Databases, Rules and Apps for the Italian Language. Ph. D. Thesis

The Sintax-Semantics Interface Il sistema LEG-Semantic Role Labeling (LEG-SRL) per l’italiano Elia, A. (2014 b). Operatori, argomenti e il sistema leg-semantic role labelling dell’italiano. In Mirto I. , a cura di, Le relazioni irresistibili, pages 105– 118. Edizioni Ets.