| Nederhof, M.-J.: 1998, Context-free parsing through regular approximation, Proceedings of the International Workshop on Finite State Methods in Natural Language Processing, Ankara, Turkey, pp. 13--24. |
.... c) 2 Sigma Theta K : X; 0) w; c)g Context free grammars used in spoken dialogue applications often represent regular languages [Church 1983, Brown and Buntschuh 1994] either by construction or as a result of a finite state approximation of a more general CFG [Pereira and Wright 1991, Nederhof 1998, Mohri and Nederhof 2000] This is because in most real time applications, general CFGs are computationally too demanding. The GRM library provides general tools for compiling a weighted CFG into a weighted finite automaton under certain general conditions described below. In addition, our ....
Nederhof, M.-J.: 1998, Context-free parsing through regular approximation, Proceedings of the International Workshop on Finite State Methods in Natural Language Processing, Ankara, Turkey, pp. 13--24.
....for the purpose of interpretation. In such a set up the finite state approximation is used as a quick filter to rule out (possibly many) impossible analyses. For a few succeeding analyses the original grammar is then applied to obtain interpretations. Algorithms for Linguistic Processing 29 In Nederhof (1998) this approach is extended by letting the finite state approximation produce a finite state transducer rather than a recogniser. The transducer produces a table, in linear time, which can then be used (in a second phase) to recover all parse trees with respect to the original grammar. The ....
Nederhof, Mark-Jan. 1998. Context-free parsing through regular approximation.
....of recognition of strings generated by X in the original grammar. This fact can be used to compile the transformed grammar into a finite state transducer that outputs bracketed strings equivalent to parse trees. The resulting parse trees retain much of the structure of the original grammar. See Nederhof [1998] for a related idea. The language generated by the transformed grammar is a superset of that of the original grammar. Indeed, it is clear from the way a rule in the original grammar is split into several rules, that any string accepted by a series of derivations in the input grammar is also ....
Nederhof, M.-J.: 1998, Context-free parsing through regular approximation, Proceedings of the International Workshop on Finite State Methods in Natural Language Processing, Ankara, Turkey, pp. 13--24.
Online articles have much greater impact More about CiteSeer.IST Add search form to your site Submit documents Feedback
CiteSeer.IST - Copyright Penn State and NEC