Marc Vilain and David Palmer. Transformation-based bracketing: Fast algorithms and experimental results. In John Carroll, editor, Workshop on Robust Parsing (ESSLLI '96), pages 93--102, 1996.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....also been applied to parsing. Brill [15] proposes starting with a uniformly right branching parse and learning rules for rotating local trees in order to improve the fit to a training corpus. Learning can be timeconsuming, but once the rules have been learned, parsing is very fast. Vilain Palmer [82] explore techniques for improving learning speeds, and mention a fast parser implementation. Voutilainen [50] describes a partial parser, ENGCG, that is very similar in operation to the constraint grammar tagger. Lexical and morphological analysis assigns a set of possible syntactic function ....

Marc Vilain and David Palmer. Transformation-based bracketing: Fast algorithms and experimental results. In John Carroll, editor, Workshop on Robust Parsing (ESSLLI '96), pages 93--102, 1996.

....dialogues of spontaneous speech (approx. 3 million words in total) see (Godfrey et al. 1992) As input for my system, I use SWB transcripts and speech recognizer hypotheses, extracted from Nbest lists (see the Appendix for an example) The latter come from a previous Switchboard evaluation (March 1996: Switchboard and Callhome databases) 3.1.4 Output The output of the system is a first best list which is derived from the re ranked Nbest lists (for every utterance) 3.1.5 Resources The Nbest lists are generated from word lattices produced by the JANUS speech recognizer (Waibel et al. ....

....in WER over size of Nbest list 4.2 Properties of the Data 4.2. 1 Data Used for General System Development For general system development, i.e. training of the neural network, testing of various system parameters, and stepwise refinement and improvement of the system components, a subset of the March 1996 Switchboard Evaluation Data was used, comprising 374 utterances in total, partly from Switchboard, partly from Callhome data. The properties of this data are given in Table 4.1; Figure 4.1 shows the potential decrease in WER if one knew which hypothesis to rank first over the length of the Nbest ....

Marc Vilain and David Palmer. 1996. Transformation-based bracketing: Fast algorithms and experimental results. In Workshop on Robust Parsing, 8th European Summer School in Logic, Language and Information, Prague, Czech Republic, pages 93--102.

....also been applied to parsing. Brill [15] proposes starting with a uniformly right branching parse and learning rules for rotating local trees in order to improve the fit to a training corpus. Learning can be timeconsuming, but once the rules have been learned, parsing is very fast. Vilain Palmer [82] explore techniques for improving learning speeds, and mention a fast parser implementation. 3 Voutilainen [50] describes a partial parser, ENGCG, that is very similar in operation to the constraint grammar tagger. Lexical and morphological analysis assigns a set of possible syntactic function ....

Marc Vilain and David Palmer. Transformation-based bracketing: Fast algorithms and experimental results. In John Carroll, editor, Workshop on Robust Parsing (ESSLLI '96), pages 93--102, 1996. 22

....knowledge is captured in opaque parameter les, typically many megabytes large. This makes it a challenge to capitalize on human intuitions to improve the machine derived grammars. An alternative to these statistical induction methods is a method called Transformation Based Parsing (TBP) [2, 11, 12]. In TBP, a grammar consists of an ordered sequence of tree transform rules. To learn the rules, we begin with some initial annotation of the training corpus (for instance, every sentence parsed as a at structure under an S node) and then we iteratively search for the transform rule whose ....

Marc Vilain and David Palmer. Transformation-based bracketing: fast algorithms and experimental results. In Proceedings of the Workshop on Robust Parsing (at ESSLLI-96), 1996.

....this, we will now turn our attention to making greater use of the techniques and technology developed for our MUC system. These components were designed to be highly efficient, very robust, and trainable for new types of text and new languages, as well as new domains and application areas [4, 5]. They include part of speech taggers and phrase and name finders, as well as higherlevel language processing components that can, for example, automatically merge multiple mentions of the same entity (e.g. John Q. Public, Mr. Public, and John) We have used this technology in the context of ....

Marc B. Vilain and David D. Palmer. "Transformation-based bracketing: fast algorithms and experimental results", Proceedings of the Workshop on Robust Parsing. 1996.

.... learning algorithm also considers the entire training set at all learning steps, rather than decreasing the size of the training data as learning progresses, such as is the case in decision tree induction (Quinlan, 1986) For a thorough discussion of transformation based learning, see Ramshaw and Marcus (1996). Brill s work provides a proof of viability of transformation based techniques in the form of a number of processors, including a (widelydistributed) part of speech tagger (Brill, 1994) a procedure for prepositional phrase attachment (Brill and Resnik, 1994) and a bracketing parser (Brill, ....

Marc Vilain and David Palmer. 1996. Transformation-based bracketing: Fast algorithms and experimental results. In Proceedings of the Workshop on Robust Parsing, held at ESSLLI 1996.

No context found.

Marc Vilain and David Palmer. 1996. Transformation-Based Bracketing: Fast Algorithms and Experimental Results. In: Workshop on Robust Parsing, ESSLLI'96.

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