A Proof System for Tree Adjoining Grammars
نویسنده
چکیده
Many TAG-based systems employ a particular tree adjoining grammar to generate the intended structures of the set of sentences they aim to describe. However, in most cases, the underlying set of elementary trees is more or less hand-made or maybe derived from a given tree data-base. We present a formal framework that allow to specify tree adjoining grammars by logical formulae. Based on this formalism we can check whether a given specification is TAG-consistent or whether a given TAG meets some particular properties. In addition, we sketch a method that generates a TAG from a given logical specification. As formal foundation, we employ a particular version of modal hybrid logic to specify the properties of T/D-trees. Such trees structurally combine a derived TAG-tree T and its associated derivation tree D. Finally, we sketch a labeled tableau calculus that constructs a set of tree automata representing the elementary trees of the specified TAG and a special tree automaton for the corresponding derivation trees. In literature, we find some approaches specifying TAGs, or more generally, mildly context-sensitive grammar formalisms, that gradually vary in their underlying framework. Commonly, either starts with a logical description of recognizable sets of trees (Thatcher and Wright, 1968). However, they differ in their method of leaving the context-free paradigm. The approach mentioned in (Morawietz and Mönnich, 2001) and (Michaelis, Mönnich and Morawietz, 2000) uses a ‘lifting’ function that encodes a TAG into a regular tree grammar. In (Rogers, 1999) (and related works) we find a logical description of TAGs that is based on a 3-dimensional view of trees. The important issue of this approach is to combine the derived TAG-tree and its derivation tree to a single 3-dimensional structure. Similarly, we also consider the derived TAG-tree and its derivation tree employ so-called T/D-trees. However we only associate the nodes of the derived tree with the corresponding node in the derivation tree. Consequently, all nodes of the same instance of an elementary tree refer to the same corresponding node in the derived tree. Therefore, we can specify structural properties of the derived TAG-tree and of the derivation tree at the same time. Using the links to the derivation tree, we can identify nodes in the TAG tree that belong to the same instance of some elementary tree. In contrast to the other approaches mentioned above which encode the TAG-tree into other kind of structures, we keep the original derived TAG tree as a structural unit. Consequently, we can directly access the nodes and the structural properties of the TAG tree without employing a particular projection function or any other special coding issues. In essence, our formalism employs modal hybrid logic that combines the simplicity of modal logic and the expressivity of classical logic. The use of so-called nominals in hybrid logic offer explicit references to certain tree nodes which is (directly) possible in modal approaches. We introduce the hybrid language HLTAG that specifies properties of the combined structure of derived TAG-trees and their derivation trees. Using this language we specify a number of TAG axioms which establish a notion of TAG-consistency. Further, we briefly illustrate a formalism that constructs a number of tree automata representing the underlying TAG for a given TAG-consistent HLTAG formula.
منابع مشابه
A Direct Link between Tree-Adjoining and Context-Free Tree Grammars
The tree languages of tree-adjoining grammars are precisely those of linear monadic context-free tree grammars. Unlike the original proof, we present a direct transformation of a tree-adjoining grammar into an equivalent linear monadic context-free tree grammar.
متن کاملPreRkTAG: Prediction of RNA Knotted Structures Using Tree Adjoining Grammars
Background: RNA molecules play many important regulatory, catalytic and structural <span style="font-variant: normal; font-style: norma...
متن کاملThe Equivalence of Tree Adjoining Grammars and Monadic Linear Context-Free Tree Grammars
The equivalence of leaf languages of tree adjoining grammars and monadic linear context-free grammars was shown about a decade ago. This paper presents a proof of the strong equivalence of these grammar formalisms. Non-strict tree adjoining grammars and monadic linear context-free grammars define the same class of tree languages. We also present a logical characterisation of this tree language ...
متن کاملUsing Feature Logic Programming to Describe Tree Transducers and Tree Adjoining Grammars
This draft paper contains a new proposal for modeling tree transductions and tree adjoining grammars using feature logic. Originally the intent was to have a declarative logic which was less complex then monadic second order logic, in which both transductions and adjunction sequences could be expressed. I feel that the feature logic descriptions are in fact somewhat easier than their monadic se...
متن کاملXTAG - A Graphical Workbench for Developing Tree-Adjoining Grammars
We describe a workbench (XTAG) for the development of tree-adjoining grammars and their parsers, and discuss some issues that arise in the design of the graphical interface. Contrary to string rewriting grammars generating trees, the elementary objects manipulated by a tree-adjoining grammar are extended trees (i.e. trees of depth one or more) which capture syntactic information of lexical item...
متن کاملExtraction of Tree Adjoining Grammars from a Treebank for Korean
We present the implementation of a system which extracts not only lexicalized grammars but also feature-based lexicalized grammars from Korean Sejong Treebank. We report on some practical experiments where we extract TAG grammars and tree schemata. Above all, full-scale syntactic tags and well-formed morphological analysis in Sejong Treebank allow us to extract syntactic features. In addition, ...
متن کامل