Defending that Grammaticality can be explained through a Fuzzy - - PowerPoint PPT Presentation
LACOMPLING 2018 Symposium on Logic and Algorithms in Computational Linguistics 2018 Stockholm, 28 31 August 2018 A DRI T ORRENS U RRUTIA D EPARTAMENT DE F ILOLOGIES R OMNIQUES U NIVERSITAT R OVIRA I V IRGILI T ARRAGONA Defending that
LACOMPLING 2018 Symposium on Logic and Algorithms in Computational Linguistics 2018
Stockholm, 28 –31 August 2018
DEPARTAMENT DE FILOLOGIES ROMÀNIQUES UNIVERSITAT ROVIRA I VIRGILI TARRAGONA
Grammaticality as an Uncertain/Fuzzy value
Tools for taking into account Grammaticality Fuzzy Grammar as a model to explain Grammaticality in terms of Degrees
“To give up the notion that a grammar defines a set of well-formed utterances is to give up a great deal […] If we can maintain the concept of discrete grammaticality, we will be in a better position to pursue an understanding of grammatical universals” Bever (1975:601). Sobre la idealización del lenguaje “The only reasonable way to approach a grasp
Natural Natural Language Language Processing Processing Do Do we we evaluate evaluate non non-
grammatical inputs? inputs?
input
PERFECT
+ Deep +Closer to thinking
Natural Natural Language Language Processing Processing We We Do Do Evaluate Evaluate Inputs Inputs
X---X---- X—X-------- X
input
This guy doesn’t speak very well
X-------X— X--------X input
OBJECTIVE: OUR MACHINES HAVE TO BOTH UNDERSTAND/PARSE AND EVALUATE THE NATURAL LANGUAGE INPUTS, AS HUMANS DO INPUT: GRAMMATICALITY AT 85% SELF-TAUGHT LANGUAGE LEARNING SOFTWARES
Phonetics Morphology Pragmatics LINGUISTIC COMPETENCE Syntax Semantics Prosody
Spanish
THIS WORK IS ASSEMBLED ON
GRAMMATICALITY
Syntax
GRADIENT EVALUATIVE SYSTEMS
CLASSICAL LOGIC <1,0> FUZZY LOGIC: [1,0]
AN INPUT IS EITHER GRAMMATICAL O NON- GRAMMATICAL LINGUISTIC INPUT DISCRETE GRAMMAR: FUZZY GRAMMAR: AN INPUT IS % GRAMMATICAL LINGUISTIC INPUT LINGUISTIC INPUT
IF THE VALUE OF GRAMMATICALITY IS HIGH, THEN THE VALUE OF NON-GRAMMATICALITY IS LOW
FTT Vilem Novak, (2005) The basic concept in FTT is type (denoted by Greek letters)
Atomic types are representing elements Type o (omicron) is the type of truth degree In the semantics In the semantics the type is assigned a set M whose elements can be anything In the semantics it is A set of truth values Mo Which in our case is Mo = [0,1]
a [0,1] Representing various degrees, Grammaticality, Complexity, etc.
Membership Function
Set
Linguistic Knowledge of a Group: Mo [0,1]
Complex types, set of functions Mo
Linguistic Knowledge of a Group:
Syntactic Module in a FG
Set of Rules
Syntactic Knowledge in FG: Mo [0,1]
Syntax in FGr
Set of Rules
The Constraints define the linguistic relation between POS and Syntax (or other Modules)
Blache, 2016
and Syntax (or other Modules) Precedence A > B Requirement A B Exclusion A B
PROPERTIES IN LINGUISTIC CONSTRUCTIONS
CONSTRUCTION IS OUR FUZZY SET
Each Category trigger a Set of Properties A Construction is Triggered by a Set of Categories PROPERTY GRAMMARS CONSTRUCTION= Set of Properties+ Set of Categories PROPER NOUN PROPN DET (Set of Properties of the PROPN) Manchester vs El Manchester (?) NOMINAL PHRASE SUBJECT DIRECT OBJECT MODIFICATOR PHRASE etc.
PROPERTIES IN LINGUISTIC CONSTRUCTIONS
DET PROPERTIE’S DET > N DET dep. N DET agree N DET Pron, PROPN ADJ PROPERTIE’S ADJ > N ADJ dep N NOUN PROPERTIE’S N DET
“The book red”
Initial Set:{Det1 N2 Adj3}
A= {{Det1 N2} {Det1 N2 Adj3}} A= {{Det1 N2} {Det1 N2 Adj3}}
Assignation Properties
{Det1 N2} P+= {Det<N; N Det; Uniq (Det, N); Oblig (N)} P-= ∅ {Det1 N2 Adj3} P+= {Det<N; N Det; Uniq (Det, N,Adj); Oblig (N) Adj
mod N}
P-= {Adj<N}
Transitivity Verb Construction In Spanish
Constraint behaviour in Fuzzy Grammar () Canonical (Gold Standard) () Violated ()Variability
In Spanish Canonical () Verb dep Direct Object (N PRON) Verb dep Subject (N PRON) Variability () 1: VS: S in verbal person (morpheme)
Linguistic Knowledge in FG: Mo [0,1]
input
ACCEPTABILITY
input
Every Rule in a Dialect (D) triggers rules in a Module of a grammar (M / X) , Both have a degree of Grammaticality
Rule1, Rule2, Rule3, Rule 4 X Is an example of rules that define the syntax of our FGr Rulea, Ruleb, Rulec, Ruled D Is an example of rules that define an input in a Dialect
X(Rule3, Rulec) = 0.6 & X(Rule4, Rulec) = 0.9 They are an example of how a rule in an input of dialect can trigger two rules in the syntax rule set of a FGr (the canonical and the variable)
THE LESS GRAMMATICALITY THE MORE RULES ARE TRIGGERED
Constraint behaviour in Fuzzy Grammar () Canonical (Gold Standard) () Violated
() Violated ()Variability
FFI2015-69978-P, Ministerio de Economía y Competitividad: “GRAMMATICAL INFERENCE ALGORITHMS
FOR MEASURING THE RELATIVE COMPLEXITY OF NATURAL LANGUAGE”
adria.torrens@urv.cat