Collins, M. (1997). Three Generative, Lexicalised Models for Statistical Parsing. In Proceedings of ACL-1997.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....In an effort to improve question focus recognition this year, we trained the Brill part ofspeech tagger [2] on questions from TREC 8, TREC 9 and HowStuffWorks. 7] The resulting rules were used to tag the TREC 10 questions. The tagged questions were then run through the Collins parser [3] [4] for a full parse. There are three steps to question focus assignment. In the first step, the question type is determined using predefined search patterns based on regular expressions. There are 7 special question types (acronym, counterpart, definition, famous, standfor, synonym, why) and 7 ....

Collins, Michael. (1997). Three Generative, Lexicalised Models for Statistical Parsing. Proceedings of the 35 Annual Meeting of the ACL (jointly with the 8th Conference of the EACL), Madrid. Available at <http://www.research.att.com/~mcollins/papers/paper14.short.ps>.

....intensive research. An interesting feature common to most such models is the incorporation of contextual dependencies on individual head words into rule based probability models. Such word based lexicalizations of probability models are used successfully in the statistical parsing models of, e.g. Collins (1997), Charniak (1997) or Ratnaparkhi (1997) However, it is still an open question which kind of lexicalization, e.g. statistics on individual words or statistics based upon word classes, is the best choice. Secondly, these approaches have in common the fact that the probability models are trained ....

Michael Collins. 1997. Three generative, lexicalised models for statistical parsing. In Proceedings of the 35th ACL, Madrid.

....1993] In this kind of work, morphological analysis has been of minor concern: actual word occurrences and their POS tags, instead of morphemes, are utilized in probabilistic language models. Impoverished forms of morphological analysis are often consulted in order to deal with unknown words (see [Collins, 1997, Charniak, 1999] Clearly, statistics over word occurrence is more prone to sparse data problems than morpheme occurrence. However, for English, the existence of large syntactic tree banks, extended POS 5 tag sets and the less complex nature of English word structure, have been instrumental in ....

....an important source for statistics over linguistic phenomena, which can be employed for resolving ambiguities during processing. In probabilistic syntactic parsing in particular, a tree bank is used for inducing probabilistic grammars e.g. Scha, 1990, Bod, 1992, Magerman, 1995, Bod, 1995, Collins, 1997, Charniak, 1999, Sima an, 2000] A probabilistic language model consists of a probabilistic grammar and a model of how probabilities of parse trees and sentences are derived. In a probabilistic grammar, a formal grammar is extended with a nite set of conditional probabilities, each associated ....

[Article contains additional citation context not shown here]

Collins, M. (1997). Three generative, lexicalized models for statistical parsing. In Proceedings of the 35th Annual Meeting of the ACL and the 8th Conference of the EACL, pages 16-23, Madrid, Spain.

....the Tree gram model, beyond those underlying DOP, is that a tree bank of analysed utterances, i.e. parse trees, can be viewed di erently if we abandon the atomicity of linguistically motivated CFG rules. This has been exploited in the statistical parsing literature [Eisner, 1996, Charniak, 1997a, Collins, 1997, Charniak, 1999] for parsing with head driven bilexical dependencies in the so called Markov Grammars 1 . In the Treegram model, we generalize over Markov Grammars as well as the DOP model by extracting constructs from the tree bank trees that abide neither by the atomicity of CFG rules nor by ....

....and with surrounding syntactic information. Indeed, as Charniak [Charniak, 1997b] observes, lexicalization with actual words is the biggest change in statistical parsing over the last few years. Head lexicalization currently pervades in the parsing literature e.g. Eisner, 1996, Collins, 1996, Collins, 1997, Charniak, 1997a, Charniak, 1999, Ratnaparkhi, 1997] This method extends every treebank nonterminal with its head word: the model is trained on this head lexicalized treebank. Head lexicalized models extract probabilistic relations between pairs of lexicalized nonterminals ( bilexical ....

[Article contains additional citation context not shown here]

Collins, M. (1997). Three generative, lexicalized models for statistical parsing. In Proceedings of the 35th Annual Meeting of the ACL and the 8th Conference of the EACL, pages 16-23, Madrid, Spain.

.... often perform as well as other broad coverage parsing systems for predicting tree structure from POS tags (Charniak, 1996) In addition, many more sophisticated parsing models are elaborations of such PCFG models, so understanding the properties of PCFGs is likely to be useful (Charniak, 1997; Collins, 1997). It is well known that natural language exhibits dependencies that Context Free Grammars (CFGs) and hence PCFGs, cannot describe (Shieber, 1985) But as explained below, the independence assumptions implicit in PCFGs introduce biases in the statistical model induced from a tree bank even in ....

Collins, Michael. 1997. Three generative, lexicalised models for statistical parsing. In The Proceedings of the 35th Annual Meeting of the Association for Computational Linguistics, San Francisco. Morgan Kaufmann.

....shallow syntactic information can be extracted using local information by examining the pattern itself, its nearby context and the local part of speech information. Thus, over the past few years, along with advances in the use of learning and statistical methods for acquisition of full parsers (Collins, 1997; Charniak, 1997a; Charniak, 1997b; Ratnaparkhi, 1997) signi cant progress has been made on the use of statistical learning methods to recognize shallow parsing patterns syntactic phrases or words that participate in a syntactic relationship (Church, 1988; Ramshaw and Marcus, 1995; Argamon et ....

....was the desire to get better performance and higher reliability. However, since work in this direction has started, a signi cant progress has also been made in the research on statistical learning of full parsers, both in terms of accuracy and processing time (Charniak, 1997b; Charniak, 1997a; Collins, 1997; Ratnaparkhi, 1997) This paper investigates the question of whether work on shallow parsing is worthwhile. That is, we attempt to evaluate quantitatively the intuitions described above that learning to perform shallow parsing could be more accurate and more robust than learning to generate ....

[Article contains additional citation context not shown here]

M. Collins. 1997. Three generative, lexicalised models for statistical parsing. In Proceedings of the 35th Annual Meeting of the Association for Computational Linguistics.

....memorizes the training data. It is well known that the one count parameters of an n gram model significantly improve the model the one count parameters significantly reduce cross entropy of the model. More sophisticated language models, such as stochastic grammars used in open domain parsing (Collins 1997; Charniak 2000) also involve a number of one count parameters essentially equal to the size of the training data. The fundamental theoretical challenge is to explain why such one count models do not overfit. Bayesian explanations for the performance of one count models can be given with an ....

Collins, M. (1997). Three generative, lexicalised models for statistical parsing. In Proceedings of the 35th Annual Meeting of the ACL.

....as described below. The first three counts (trans, pass, vbn) were performed on the LDC s 65 million word tagged ACL DCI corpus (Brown, and Wall Street Journal 1987 1989) The last two counts (caus and anim) were performed on a 29 million word parsed corpus (Wall Street Journal 1988, provided by Michael Collins (Collins, 1997)) The features were counted as follows: trans: The closest noun following a verb was considered a potential object. A verb immediately followed by a potential object was counted as transitive, otherwise as intransitive. pass: A token tagged VBD (the tag for simple past) was counted as active. A ....

Michael John Collins. 1997. Three generative, lexicalised models for statistical parsing. In Procs of ACL'97, pages 16--23. Madrid, Spain.

.... A very promising idea, spelt out in some detail in Langkilde s (1999) thesis proposal is to elaborate both the symbolic generator and the statistical extractor to work with parse forests in the format of the Penn Treebank. The concrete proposal for a statistical extractor is to interpolate Collins (1997) dependency model with a bigram model, using techniques recently proposed by Chelba Jelinek (1998) Concomitant changes in the symbolic generator allow it Article submitted to Royal Society Stochastic text generation 7 to produce treebank style trees instead of at strings. To the extent that ....

Collins, M. 1997. Three generative, lexicalised models for statistical parsing. In Proceedings of the 35th Annual Meeting of the ACL (jointly with the 8th Conference of the EACL), Madrid.

....as described below. The #rst three counts #trans, pass, vbn#were performed on the LDC s 65 million word tagged ACL#DCI corpus #Brown, and Wall Street Journal 1987#1989#. The last two counts #caus and anim#were performed on a 29 million word parsed corpus #Wall Street Journal 1988, provided by Michael Collins #Collins, 1997##. The features were counted as follows: trans: The closest noun following a verb was considered a potential object. Averb immediately followed by a potential object was counted as transitive, otherwise as intransitive. pass: A token tagged VBD #the tag for simple past# was counted as active. ....

Michael John Collins. 1997. Three generative, lexicalised models for statistical parsing. In Procs of ACL'97, pages 16#23. Madrid, Spain.

....see the man with the 1 Our main motivation is that in observing the behavior of Nitrogen s word pair based statistical ranking, we notice many errors due to missed long distance dependencies. We believe that many of these errors will be corrected if we use a syntax based ranking, such as that of Collins (1997) or Chelba and Jelinek (1998) operating over trees rather than flat strings. In order to match up with statistically collected data, our trees are compliant with the labeling and bracketing scheme of the Penn Treebank (Marcus, Santorini, and Marcinkiewicz, 1993) Another advantage of using ....

Collins, M. (1997). Three Generative, Lexicalised Models for Statistical Parsing, Proc. ACL.

....It presents a new model based on structural relations, the Tree gram model, and reports experiments showing that structural relations should bene t from enrichment by lexical dependencies. 1 Introduction Head lexicalization currently pervades in the parsing literature e.g. Eisner, 1996; Collins, 1997; Charniak, 1999) This method extends every treebank nonterminal with its headword: the model is trained on this head lexicalized treebank. Head lexicalized models extract probabilistic relations between pairs of lexicalized nonterminals ( bilexical dependencies ) every relation is between a ....

....a POS tag) a pre head representing its head word. The pre head of node is extracted from the constituent parse tree under node . In this paper, the pre head of consists of 1) the POS tag of the head word of (called 1 st order pre heads or 1 PH ) and 1 Head identi cation procedure by (Collins, 1997). possibly 2) the label of the mother node of that POS tag (called 2 nd order or 2 PH ) Preheads here also include other information de ned in the sequel, e.g. subcat frames. The complex categories that result from the enrichment serve as the nonterminals of our training treebank; we refer ....

[Article contains additional citation context not shown here]

M. Collins. 1997. Three generative, lexicalized models for statistical parsing. In Proceedings of the 35th Annual Meeting of the ACL and the 8th Conference of the EACL, pages 16-23, Madrid, Spain.

....suggests that stable cross corpus subcategorization frequencies may be found when verb sense is adequately controlled. Introduction Verb subcategorization probabilities play an important role in both computational linguistic applications (e.g. Carroll, Minnen, and Briscoe 1998, Charniak 1997, Collins 1996 1997, Joshi and Srinivas 1994, Kim, Srinivas, and Trueswell 1997, Stolcke et al. 1997) and psycholinguistic models of language processing (e.g. Boland 1997, Clifton et al. 1984, Ferreira McClure 1997, Fodor 1978, Garnsey et al. 1997, Jurafsky 1996, MacDonald 1994, Mitchell Holmes 1985, Tanenhaus ....

Collins, M. J. (1997) Three generative, lexicalised models for statistical parsing. In Proceedings of ACL-97.

....since the late 1980 s. The success of such approaches in areas like speech recognition (Rabiner, 1989; Bahl, Jelinek, Mercer, 1983) part of speech tagging (Charniak, Hendrickson, Jacobson, Perkowitz, 1993) syntactic parsing (Ratnaparkhi, 1999; Manning Carpenter, 1997; Charniak, 1996; Collins, 1997; Pereira Shabes, 1992) and text or discourse segmentation (Litman, 1996) is evidential. In fact, it has been coined a revolution within the NLP community (Hirschberg, 1998) There are reasons why such approaches have experienced a resurgence: 1) the success in information and networking ....

Collins, M. J. (1997). Three generative, lexicalised models for statistical parsing. In Proceedings of the 35th Annual Meeting of the Association for Computational Linguistics, pp. 16--23.

....for instance, Earley s parsing method can be adopted for stochastic parsing (see [Stolke, 1993] However this model su ers greatly because of data sparseness: there are not enough examples of productions A seen in the training set so that the probabilities can be reliably estimated. Collins ([Collins, 1997]) avoids this problem by rst determining the head phrase of the right hand side of any production A X 1 : X k , A 2 N;X i 2 T [ N(in a similar fashion to [Jelinek et al. 1994] and [Magerman, 1995] and using this head word to predict the words to the left and the words to the right of it. ....

Collins, M. J. (1997). Three generative, lexicalized models for statistical parsing. In Proceedings of the 35th Meeting of Association for Computational Linguistics.

....problematic, because the corpora are different and are sometimes not even described in other work. In most cases Korean research groups also use other evaluation metrics, particularly dependency accuracy, which is often used in dependency structure approaches. Training on about 40,000 sentences (Collins, 1997) achieves a crossing brackets rate of 1.07, a better value than our 1.63 value for regular parsing or the 1.13 value assuming perfect segmentation tagging, but even for similar text types, comparisons across languages are of course problematic. It is clear to us that with more training sentences, ....

M. J. Collins. 1997. Three Generative, Lexicalised Models for Statistical Parsing. In 35th Proceedings of the ACL, pages 16--23.

....word of text. The module is also able to infer some of the information represented by the Treebank II formalism even if this information has not been explicitly coded in its input. For instance, it can still analyze parse trees produced by statistical parsers such as the one developed by Collins (Collins, 1997), the output of which lacks some of the semantic cues that have been added to the Treebank II manually by annotators. To evaluate the accuracy of the automatic predicateargument analyzer, we examined 65 sentences from our sense tagged running corpus containing 162 automatically annotated verb ....

Collins, M., 1997. Three generative, lexicalised models for statistical parsing. In Proceedings of the 35th Annual Meeting of the Association for Computational Linguistics.

....very simple source and channel models. In a departure from the above discussion and from previous work on statistical channel models, we assign probabilities P tree (s) and P expand tree (t j s) to trees rather than strings. In decoding a new string, we first parse it into a large tree t (using Collins parser (1997)) and we then hypothesize and rank various small trees. Good source strings are ones that have both (1) a normal looking parse tree, and (2) normal looking word pairs. P tree (s) is a combination of a standard probabilistic context free grammar (PCFG) score, which is computed over the grammar ....

....yielded an accuracy of 87.16 ( Sigma 0.14) A majority baseline classifier that chooses the action shift has an accuracy of 28.72 . Employing the decision based model To compress sentences, we apply the shift reduce drop model in a deterministic fashion. We parse the sentence to be compressed (Collins 1997) and we initialize the input list with the words in the sentence and the syntactic constituents that begin at each word, as shown in Figure 4. We then incrementally inquire the learned classifier what action to perform, and we simulate the execution of that action. The procedure ends when the ....

Collins, M. 1997. Three generative, lexicalized models for statistical parsing. In Proceedings of the 35th Annual Meeting of the Association for Computational Linguistics (ACL--97), 16--23.

....broadly applicable. 1.1 History Based Models Before discussing the reranking approaches, we will describe history based models (Black et al. 1992) They are important for a few reasons. First, at present the best performing parsers on the WSJ treebank (Ratnaparkhi 1997; Charniak 1997, 1999; Collins 1997, 1999) are all cases of history based models. Many systems applied to part ofspeech tagging, speech recognition and other language or speech tasks also fall into this class of model. Second, a particular history based model (that of Collins (1999) will be used as the initial model for our ....

Collins, M. (1997). Three generative, lexicalised models for statistical parsing. In Proceedings of the 35th Annual Meeting of the Association for Computational Linguistics and 8th Conference of the European Chapter of the Association for Computational Linguistics (pp. 16--23). San Francisco: Morgan Kaufmann.

No context found.

Collins, M. (1997). Three Generative, Lexicalised Models for Statistical Parsing. In Proceedings of ACL-1997.

No context found.

M. J. Collins. 1997. Three generative, lexicalised models for statistical parsing. In Proceedings of the 35th Annual Meeting of the ACL, Madrid. Association for Computational Linguistics.

No context found.

Collins, M.J. 1997. Three generative, lexicalised models for statistical parsing. In The Proceedings of the 35th Annual Meeting of the Association for Computational Linguistics, pages 16--23.

No context found.

Collins, M. 1997. Three generative, lexicalized models for statistical parsing. Proceedings of the 35th Annual Meeting of the Association for Computational Linguistics (pp. 16-23). Madrid, Spain.

No context found.

Collins, M. 1997. Three generative, lexicalised models for statistical parsing. In Proc. Assoc. for Comp. Linguistics, 16--23.

No context found.

Michael Collins. 1997. Three generative, lexicalized models for statistical parsing. In proceedings of the ACL97, Madrid, Spain.

No context found.

M. Collins. 1997. Three generative, lexicalised models for statistical parsing. In Proceedings of the 35th Annual Meeting of the Association for Computational Linguistic, Madrid, Spain.

No context found.

Collins, Michael. 1997. Three generative, lexicalised models for statistical parsing. In Proceedings of the 35th Annual Conference of the Association for Computational Linguistics (ACL-97), pages 16--23, Madrid, Spain.

No context found.

Michael Collins. 1997. Three Generative, Lexicalized Models for Statistical Parsing. In 35th Annual Meeting of the ACL.

No context found.

Collins, Michael. (1997). Three Generative, Lexicalised Models for Statistical Parsing. In Proceedings of the 35th Annual Meeting of the Association for Computational Linguistics and 8th Conference of the European Chapter of the Association for Computational Linguistics, pages 16-23.

No context found.

Collins, M. 1997. Three generative, lexicalized models for statistical parsing. Proceedings of the 35th Annual Meeting of the Association for Computational Linguistics (pp. 16-23). Madrid, Spain.

No context found.

Michael Collins. 1997. Three generative, lexicalised models for statistical parsing. In Proc. of ACL-97.

No context found.

Michael Collins. 1997. Three generative, lexicalized models for statistical parsing. In Proceedings of the ACL'97,Somerset, New Jersey.

No context found.

M. Collins. 1997. Three generative, lexicalised models for statistical parsing Proceedings of 35th Annual Meeting of the ACL, Madrid, Spain.

No context found.

Collins, M. (1997). Three generative, lexicalized models for statistical parsing. In Proceedings of the 35th Annual Meeting of the ACL (jointly with the 8th Conference of the EACL), Madrid.

No context found.

M. J. Collins. 1997. Three generative, lexicalised models for statistical parsing. In Proceedings of the 35th Annual Meeting of the ACL, Madrid. Association for Computational Linguistics.

No context found.

Michael John Collins. 1997. Three generative, lexicalised models for statistical parsing. In ACL 35/EACL 8, pages 16--23.

No context found.

Michael Collins. 1997. Three generative, lexicalised models for statistical parsing. In Proceedings of the 35th Annual Meeting of the ACL.

No context found.

Michael John Collins. 1997. Three generative, lexicalised models for statistical parsing. In ACL 35/EACL 8, pages 16--23.

No context found.

Michael Collins. 1997. Three generative, lexicalised models for statistical parsing. In Proceedings of the 35th Annual Meeting of the Association for Computational Linguistics, pages 16#23.

No context found.

Michael Collins. 1997. Three generative, lexicalised models for statistical parsing. In Proceedings of the 35th Annual Meeting of the ACL, pages 16{ 23, Madrid, Spain.

No context found.

Michael Collins. 1997. Three generative, lexicalised models for statistical parsing. In Proceedings of the 35th Annual Meeting of the Association for Computational Linguistics, pages 16-23.

No context found.

Michael Collins. 1997. Three generative, lexicalised models for statistical parsing. In The Proceedings of the 35th Annual Meeting of the Association for Computational Linguistics, San Francisco. Morgan Kaufmann.

No context found.

Michael Collins. 1997. Three generative, lexicalised models for statistical parsing. In Proceedings of the Annual Meeting of the Association of Computational Linguistics, volume 35, Madrid.

No context found.

Michael Collins. 1997. Three generative, lexicalised models for statistical parsing. In Proceedings of ACL-EACL '97, pages 1623.

No context found.

Michael Collins. 1997. Three generative, lexicalised models for statistical parsing. In Proceedings of ACL-EACL '97, pages 1623.

No context found.

Michael Collins. 1997. Three generative, lexicalised models for statistical parsing. In Proceedings of the 35th Annual Meeting of the ACL.

No context found.

Michael John Collins. 1997. Three generative, lexicalised models for statistical parsing. In ACL 35/EACL 8, pages 16-23.

No context found.

Michael Collins. 1997. Three generative, lexicalised models for statistical parsing. In Proceedings of the 35th Annual Meeting of the Association for Computational Linguistics, pages 16--23.

No context found.

Michael Collins. 1997. Three generative, lexicalised models for statistical parsing. In The Proceedings of the 35th Annual Meeting of the Association for Computational Linguistics, San Francisco. Morgan Kaufmann.

No context found.

Michael Collins. 1997. Three generative, lexicalised models for statistical parsing. In Proceedings of the 35th Annual Meeting of the Association for Computational Linguistics, pages 16-23.

First 50 documents  Next 50

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