Goodman, Joshua. 1997. Probabilistic feature grammars. In Proceedings of the International Workshop on Parsing Technologies 1997.  Home/Search   Document Details and Download   Summary   Related Articles   Check

.... explored a variety of ways to improve LM performance [2, 3, 4, 5] There has been good progress in developing structured models [2, 3, 5] that focus more on the hierarchical characteristics of a language than specific information about words and their lexical features (e.g. case, number) Goodman [6] integrates a small set of features at the level of Context free Grammar (CFG) production rules to achieve increased parse accuracy; however, only a small set of lexical features can be integrated without causing a significant increase in grammar size and a concomitant data sparsity problem. In ....

J. Goodman, "Probabilistic feature grammars," in Proceedings of the Fourth International Workshop on Parsing Technologies, 1997.

....final results seems counterintuitive, and merits further investigation. Future work on the learner will be in three main di rections: Abandonment of the SCFG as the basis of the lan guage model. We axe considering either Abney s random fields [1] or Goodman s Probabilistic Feature Grammmars [14] as a replacement. Apart from performance improvements, altering the model class should allow empirical investigation of the MDL claim that model classes can be evaluated in terms of compression. So, if we discover even more compact models using (say) Goodman s scheme than we could using our SCFG, ....

Joshua Goodman. Probabilistic Feature Grammars. In 5 tt' International Workshop on Pa- ing Technologies, MIT, Cambridge, Massachusetts, USA, September 1997.

.... grammar [21] head driven phrase structure grammar [15] tree insertion grammar [19] combinatorial categorial grammar [22] and bilexical grammar [8] Probabilistic lexicalized grammars have also been exploited in state of the art, real world parsers, as reported in [12] 1] 7] 3] 5] and [10]. Other parsers or language models for speech recognition that do not directly exploit a generative grammar, still are heavily based on lexicalization, as for instance the systems presented in [11] 13] 16] and [4] The wide diffusion of lexicalized grammars is mainly due to the capability of ....

....that are common to several of the above mentioned lexicalized grammars. In what follows, we report a formal definition of LCFGs and give a brief outline of the results that will be more carefully presented in the talk. Other abstractions of lexicalized grammars have been presented in [2] and in [10]. In those works, however, the major emphasis is on the study of linguistic expressiveness and of parameter estimation problems, respectively, for different lexicalized grammar formalisms. 2 Lexicalized context free grammars We can think of a lexicalized context free grammar as a particular kind ....

J. Goodman. Probabilistic feature grammars. In Proceedings of the 5 Int. Workshop on Parsing Technologies, MIT, Cambridge, MA, September 1997.

....This nal results seems counterintuitive, and merits further investigation. Future work on the learner will be in three main directions: Abandonment of the SCFG as the basis of the language model. We are considering either Abney s random elds [1] or Goodman s Probabilistic Feature Grammmars [14] as a replacement. Apart from performance improvements, altering the model class should allow empirical investigation of the MDL claim that model classes can be evaluated in terms of compression. So, if we discover even more compact models using (say) Goodman s scheme than we could using our SCFG, ....

Joshua Goodman. Probabilistic Feature Grammars. In 5 th International Workshop on Parsing Technologies, MIT, Cambridge, Massachusetts, USA, September 1997.

....pairs of head words in conjunction with chart parsing techniques to achieve high accuracy. Collins (1996, 1997) uses chart parsing techniques with head word bigram statistics derived from a treebank. Charniak (1997) uses head word bigram statistics with a probabilistic context free grammar, while Goodman (1997) uses head word bigram statistics with a probabilistic feature grammar. Collins (1996) Goodman (1997) Charniak (1997) Collins (1997) do not use general machine learning algorithms, but instead develop specialized statistical estimation techniques for their respective parsing tasks. The parser ....

....1997) uses chart parsing techniques with head word bigram statistics derived from a treebank. Charniak (1997) uses head word bigram statistics with a probabilistic context free grammar, while Goodman (1997) uses head word bigram statistics with a probabilistic feature grammar. Collins (1996) Goodman (1997), Charniak (1997) Collins (1997) do not use general machine learning algorithms, but instead develop specialized statistical estimation techniques for their respective parsing tasks. The parser in this paper attempts to combine the advantages of other approaches. It uses a natural and direct ....

[Article contains additional citation context not shown here]

Goodman, J. (1997). Probabilistic Feature Grammars. In Proceedings of the International Workshop on Parsing Technologies.

....final results seems counterintuitive, and merits further investigation. Future work on the learner will be in three main directions: ffl Abandonment of the SCFG as the basis of the language model. We are considering either Abney s random fields [1] or Goodman s Probabilistic Feature Grammmars [14] as a replacement. Apart from performance improvements, altering the model class should allow empirical investigation of the MDL claim that model classes can be evaluated in terms of compression. So, if we discover even more compact models using (say) Goodman s scheme than we could using our SCFG, ....

Joshua Goodman. Probabilistic Feature Grammars. In 5 th International Workshop on Parsing Technologies, MIT, Cambridge, Massachusetts, USA, September 1997.

....The Lexicalized Tree Insertion Grammar formalism (LTIG) has been proposed as a way to lexicalize context free grammars (Schabes This material is based upon work supported by the National Science Foundation under Grant No. IRI 9712068. We thank Yves Schabes and Stuart Shieber for their guidance; Joshua Goodman for his PCFG code; Lillian Lee and the three anonymous reviewers for their comments on the paper. and Waters, 1994) We now apply a probabilistic variant of this formalism, Probabilistic Tree Insertion Grammars (PLTIGs) to natural language processing problems of stochastic parsing and language ....

Joshua Goodman. 1997. Probabilistic feature grammars. In Proceedings of the International Workshop on Parsing Technologies 1997.

....purposes Black et al. replace their complex featurebased categories by 50 atomic syntactic categories and 50 atomic semantic categories so there are at most 2500 possible linguistic categories available, compared to the 23,431 categories they report as actually arising in parsing their corpus. Goodman (1997) describes a method based on independence assumptions and back off smoothing that could, in principle, be applied to arbitrary unification grammars, but he tests it only on a simplified model with many fewer features than a realistic unification grammar of a natural language would need, and he ....

Goodman J. (1997) "Probabilistic Feature Grammars," in Proceedings of the International Workshop on Parsing Technologies, Boston, Massachusetts.

....conjunction with chart parsing techniques to achieve high accuracy. The parsers in [Collins, 1996, Collins, 1997] use chart parsing techniques and head word bigram statistics derived from a treebank. Charniak, 1997] uses head word bigram statistics with a probabilistic context free grammar, while [Goodman, 1997] uses head word bigram statistics with a probabilistic feature grammar. Collins, 1996, Goodman, 1997, Charniak, 1997, Collins, 1997] do not use general machine learning algorithms, but instead develop specialized statistical estimation techniques for their respective parsing tasks. The parser in ....

....1997] use chart parsing techniques and head word bigram statistics derived from a treebank. Charniak, 1997] uses head word bigram statistics with a probabilistic context free grammar, while [Goodman, 1997] uses head word bigram statistics with a probabilistic feature grammar. Collins, 1996, Goodman, 1997, Charniak, 1997, Collins, 1997] do not use general machine learning algorithms, but instead develop specialized statistical estimation techniques for their respective parsing tasks. The parser in this paper attempts to combine the advantages of other approaches. It uses a natural and direct ....

[Article contains additional citation context not shown here]

Goodman, J. (1997). Probabilistic feature grammars. In Proceedings of the International Workshop on Parsing Technologies.

....there are dependencies between phenomena treated by distinct PV Ss is at best only indirect evidence against the psychological claim that this is the appropriate cognitive model. For further discussion of probabilistic interpretation of similar representation languages see Abney (1997) and Goodman (1997). 5 Because the parameters of variation are a set of binary valued PV Ss with uniform probability assigned to unset PV Ss, the product of these PV Ss effectively defines an informative prior on G consistent with the minimum description length principle (Rissanen, 1989) A more sophisticated ....

Goodman, J. (1997) `Probabilistic feature grammars', Proceedings of the 5th Int.

No context found.

Goodman, Joshua. 1997. Probabilistic feature grammars. In Proceedings of the International Workshop on Parsing Technologies 1997.

No context found.

Biometrika, 40(3 and 4):237--264. Goodman, Joshua. 1997. Probabilistic feature grammars. In Proceedings of the International Workshop on Parsing Technologies 1997.

No context found.

Goodman, J. (1997) `Probabilistic feature grammars', Proceedings of the 5th Int.

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