Estudio comparativo del rendimiento de analizadores sintácticos para gramáticas de adjunción de árboles

In this paper, we study the behavior of some of the most popular parsing algorithms for Tree Adjoining Grammars (TAG). To this end, we apply a compilation technique allowing automatic transformation of parsing schemata into efficient implementations of their corresponding algorithms; which allows us to compare the performance of different parsers in an homogeneous environment. In our study, we analyze the different algorithms both by using artificially generated grammars, which allow us to estimate the influence of grammar size in performance, and by applying the generated implementations to a real-life, wide coverage grammar (the XTAG).