Methodological Foundations of Lexicon Building
Résumé
The present paper on lexicon modelling relates to the issue of building a generic and reusable lexicon. It is founded on the basal idea that there cannot be any adequate formalization or modelling without careful and explicit description of data. Here, data are lexical semantic data.
§ 1 sketches an overview of the domain and § 2 provides terminological and conceptual clarifications. It is noted that there is no explicit agreement on what semantic data are. A set of requirements the data must meet is defined, such as being accessible to actual and realistic observation, being intersubjective, being independent of any particular theory, etc.
§ 3 is the central chapter of the paper. It defines a conceptual specification of a constructive reusable generic lexicon. First, a series of statements on observational semantics are presented: on the paradigm (precisely that of empirical science), on the syntactic descriptions of natural language expressions (a strict and narrow definition), on the description of real actual worlds, and on the observer (the indispensable black box which turns factual observations into data descriptions). These statements are basic assumptions the reader should accept before going any further. Next, two propositions on semantic modelling are made, each with explicit assumptions on observational semantics (including the observer’s knowledge and capabilities), specific data and specific requirements for theories on these.
In the first proposition, the observer uses a set of three predicates on sentences (syntactically well-formed, constructible, deviant), yielding a specific data set. As linguistic theories are not much explicit to what their target is with respect to this data set, strong difficulties are to be expected in the specification of semantic information in the lexicon, as well as a low level of intersubjectivity of data. From the observation of a few examples, it is argued it is impossible to model any lexicon semantics within the paradigm of empirical sciences in the terms of this (commonplace) proposition.
In the second proposition, the observer has in addition the ability to see that two expressions have or haven’t an identity of interpretation. This allows to partition into two sub-classes the problematic class of sentences of the first proposition (i.e. well-formed, constructible, deviant sentences), depending on whether there is or not a well-formed, constructible, non deviant paraphrase. When all conceivable paraphrases are deviant, the problem is ontological, when a non deviant paraphrase exists, the problem is relevant to linguistics. Lexical semantics are thus distinguished by a principle from encyclopaedic semantics.
§ 1 sketches an overview of the domain and § 2 provides terminological and conceptual clarifications. It is noted that there is no explicit agreement on what semantic data are. A set of requirements the data must meet is defined, such as being accessible to actual and realistic observation, being intersubjective, being independent of any particular theory, etc.
§ 3 is the central chapter of the paper. It defines a conceptual specification of a constructive reusable generic lexicon. First, a series of statements on observational semantics are presented: on the paradigm (precisely that of empirical science), on the syntactic descriptions of natural language expressions (a strict and narrow definition), on the description of real actual worlds, and on the observer (the indispensable black box which turns factual observations into data descriptions). These statements are basic assumptions the reader should accept before going any further. Next, two propositions on semantic modelling are made, each with explicit assumptions on observational semantics (including the observer’s knowledge and capabilities), specific data and specific requirements for theories on these.
In the first proposition, the observer uses a set of three predicates on sentences (syntactically well-formed, constructible, deviant), yielding a specific data set. As linguistic theories are not much explicit to what their target is with respect to this data set, strong difficulties are to be expected in the specification of semantic information in the lexicon, as well as a low level of intersubjectivity of data. From the observation of a few examples, it is argued it is impossible to model any lexicon semantics within the paradigm of empirical sciences in the terms of this (commonplace) proposition.
In the second proposition, the observer has in addition the ability to see that two expressions have or haven’t an identity of interpretation. This allows to partition into two sub-classes the problematic class of sentences of the first proposition (i.e. well-formed, constructible, deviant sentences), depending on whether there is or not a well-formed, constructible, non deviant paraphrase. When all conceivable paraphrases are deviant, the problem is ontological, when a non deviant paraphrase exists, the problem is relevant to linguistics. Lexical semantics are thus distinguished by a principle from encyclopaedic semantics.
Fichier principal
1995_Methodological_foundations_of_lexicon_building.pdf (2.4 Mo)
Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)