Tabular algorithms for TAG parsing

Abstract

We describe several tabular algorithms for Tree Adjoining Grammar parsing, creating a continuum from simple pure bottom-up algorithms to complex predictive algorithms and showing what transformations must be applied to each one in order to obtain the next one in the continuum.

Mining conceptual graphs for knowledge acquisition

“…1 probabilistic dependencies value does not change with time. 2 probabilistic dependencies are limited to n states.…”

“…Our aim is to avoid the elimination of any parsing branch until we are sure it is not used for knowledge acquisition. In the particular case of tags, the system implements a parse scheme [1] verifying the vpp, which assures the best time behavior. The introduction of a high degree of abstraction in the design of the analyzer is achieved on the basis of the mg concept [5], by involving elementary constraints re-grouped in classes, these themselves inserted in a hierarchy of multiple heritage.…”

“…Class Word Position Class Word Position Class teinte [2] color couleur [2] color tache [2] color teinté [2] color texture [2] texture position [2] position taille [1,2] size diamètre [1] sizeépaisseur [1] size longueur [1] size hauteur [1] size largeur [1] size altitude [2] size atteindre [2] size dépassant [2] size atteindre [1] organ/fruit forme [2] form bord [1] organ/fruit Table 3: File of linguistic markers lemma is not included in the lexicon of the tagger and whose corresponding tag can only be suggested through the parser from the grammar considered. This is shown in Fig.…”

“…So, although hmms can apply on arbitrary n-grams, the value of n is fixed. This firstly implies that we are assuming stationary time 1 and limited history 2 hypothesis which, for example, would not enable us to deal with long distance semantic dependencies nor recursive structures. Although we could get round this problem by considering a sufficiently high value for n, this would involve prohibitive time and space performances.…”

Extraction of complex index terms in non-English IR: A shallow parsing based approach

“…Full parsing (Alonso et al, 1999;Sikkel, 1997) is non-viable because of its high computational cost, which makes its application on a large scale impractical. Moreover, the lack of robustness of such approaches, which seek to obtain a complete parse of the whole sentence, greatly reduces their coverage (Arampatzis et al, 2000).…”