Ratnaparkhi, A. (1996). A maximum entropy model for part-of-speech tagging. In: Proc. Conference on Empirical Methods in Natural Language Processing, 133--142, Pennsylvania.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....Tagged Words by A Total Tagged Words) 100( Several approaches to the automatic tagging of texts have been proposed. The so called stochastic strategies use various statistical models, namely Markov models (MM) Merialdo, 1994] the exponential (EXP) model and the maximum entropy (ME) model ([Ratnaparkhi, 1996]. A memory based (MB) strategy represents a kind of supervised learning based on similarity based reasoning ( Daelemans, Zavrel, 1996] In a rule based (RB) strategy, a set of meaningful rules is automatically acquired. Neural networks (NE) Schmid, 1994] represent an artificial intelligence ....

A. Ratnaparkhi. A Maximum Entropy Model for Part-of-Speech Tagging. In Proceedings of the First Empirical Methods in Natural Language Processing Conference, pp. 133-141, Philadelphia, USA, 1996.

....2.1 Text Classification Text classification, as presented here, is based on the maximum entropy technique. Maximum entropy modeling shows to be a promising approach for a large variety of tasks, e.g. language modeling [Rosenfeld 94, Martin et al. 97, Peters Klakow 99] part of speech tagging [Ratnaparkhi 96] context free parsing [Ratnaparkhi 98] and text classification [Nigam et al. 99] Maximum entropy is a general technique for modeling probability distributions. Deriving the right information from the training data makes the obtained probability distribution converge towards the optimal , ....

Adwait Ratnaparkhi, A maximum entropy model for part-of-speech tagging. In Proceedings of Conference on Empirical Methods in Natural Language Processing, University of Pennsylvania, 1996.

....as well as the claimed advantages of conditional models by evaluating HMMs, MEMMs and CRFs with identical state structure on a part of speech tagging task. 2. The Label Bias Problem Classical probabilistic automata (Paz, 1971) discriminative Markov models (Bottou, 1991) maximum entropy taggers (Ratnaparkhi, 1996), and MEMMs, as well as non probabilistic sequence tagging and segmentation models with independently trained next state classifiers (Punyakanok Roth, 2001) are all potential victims of the label bias problem. For example, Figure 1 represents a simple finite state model designed to distinguish ....

....models with the global normalization of random field models. Other applications of exponential models in sequence modeling have either attempted to build generative models (Rosenfeld, 1997) which involve a hard normalization problem, or adopted local conditional models (Berger et al. 1996; Ratnaparkhi, 1996; McCallum et al. 2000) that may suffer from label bias. Non probabilistic local decision models have also been widely used in segmentation and tagging (Brill, 1995; Roth, 1998; Abney et al. 1999) Because of the computational complexity of global training, these models are only trained to ....

Ratnaparkhi, A. (1996). A maximum entropy model for part-of-speech tagging. Proc. EMNLP. New Brunswick, New Jersey: Association for Computational Linguistics.

....cannot capture lexical information which is necessary for resolving some morphological ambiguity. Some recent works have reported that tagging accuracy could be improved by using lexical information in their models such as the transformation based patch rules(Brill, 1994) the maximum entropy model(Ratnaparkhi, 1996), the statistical lexical rules(Lee et al. 1999) the HMM considering multi words(Kim, 1996) the selectively lexicalized HMM(Kim et al. 1999) and so on. In the previous works(Kim, 1996) Kim et al. 1999) however, their HMMs were lexicalized selectively and restrictively. In this paper we ....

A. Ratnaparkhi. 1996. A Maximum Entropy Model for Part-of-Speech Tagging. In ##### ## ### ######### ####### ## ####### #### ##### ########## ####################, 133-142.

....Section 5 for details) Collins s parser does not operate directly on word sequences, however. Rather, it expects a sequence of part of speech tags T as input. It finds K that maximizes P (K ; T ) We relied on another existing tool, Adwait Ratnaparkhi s maximum entropy part of speech tagger [22] to compute the best tagging T for a given sentence S, and then applied the parser to T . Note that this represents another approximation since we are not jointly optimizing T and K. Again, we felt confident in this approximation because we were able to verify the high accuracy of the tagger on ....

....For the tagger, this is the percentage of tags that were labeled correctly. For the parser, it is customary to evaluate the precision and recall of constituents (bracket pairs) For comparison purposes, we also give the corresponding figures for the Wall Street Journal corpus, as reported in [22] and [2] respectively. As can he seen, the Switchboard performance of both tagger and parser is within 1 2 of the WSJ figures. The amount of training data for tagging was roughly the same in both cases (1 millions words) but the Switchboard parser was trained on only 226,000 words, about 1 6 ....

Adwait Ratnaparkhi. A maximum entropy model for part-of-speech tagging. In Eric Brill and Kenneth Church, editors, Proceedings of the Conference on Empirical Methods in Natural Language Processing, pages 133--142, University of Pennsylvania, Philadelphia, Pa., May 1996.

....in a mode where we informed them of the correct Treebank tags. This did not completely determine the more highly articulated tags that our system actually uses (see x4) but it did constrain the choice of tags suciently to reduce tagging error on our tag set to just 1.6 . The tagger Collins uses [Ratnaparkhi 1996] has error of about 3 on the Treebank tag set, putting Collins at a slight disadvantage; but this is mitigated somewhat by the fact that Collins trains his parser on the slightly erroneous output of the tagger rather than the correct tags (p.c. The principal bene t of feeding tags to the ....

Adwait Ratnaparkhi. 1996. A maximum-entropy model for part-of-speech tagging. Conference on Empirical Methods in Natural Language Processing.

....cannot capture lexical information which is necessary for resolving some morphological ambiguity. Some recent works have reported that tagging accuracy could be improved by using lexical information in their models such as the transformation based patch rules(Brill, 1994) the maximum entropy model(Ratnaparkhi, 1996), the statistical lexical rules(Lee et al. 1999) the HMM considering multi words(Kim, 1996) the selectively lexicalized HMM(Kim et al. 1999) and so on. In the previous works(Kim, 1996) Kim et al. 1999) however, their HMMs were lexicalized selectively and restrictively. In this paper we ....

A. Ratnaparkhi. 1996. A Maximum Entropy Model for Part-of-Speech Tagging. In Proc.

....words www for a window of size x in the y th position, stands for capitalized words. Shapire 1998) The second technique, maximum entropy modeling, has been previously applied to a variety of natural language tasks, the closest application to ours being part of speech tagging as described in (Ratnaparkhi 1996). Both of these methods learn simple weighted rules, each rule using a feature to predict one of the labels with some weight: an example rule would be, if the segment contains the n gram this is NPR news vote for label Anchor with weight 0.3. On test data examples, the label with the highest ....

....weighted vote is taken as the output of the algorithm. The boosting approach greedily searches for a subset of the features which predict the label with high accuracy; in the maximum entropy method all features occurring above a certain number of times (in our case 12) were used by the model ((Ratnaparkhi 1996) also used a count cut o# to select features) Results and Evaluation In this section we first discuss the accuracy of the method on human transcripts, focusing on the contribution of di#erent feature types to the method s performance. We then discuss results on ASR output. We divided our data ....

Ratnaparkhi, A. 1996. A maximum entropy model for part-of-speech tagging. In Proceeding of the Conference on Empirical Methods in Natural Language Processing.

No context found.

Ratnaparkhi, A. (1996). A maximum entropy model for part-of-speech tagging. In: Proc. Conference on Empirical Methods in Natural Language Processing, 133--142, Pennsylvania.

No context found.

Ratnaparkhi, Adwait, 1996, "A Maximum Entropy Model for Part-Of-Speech Tagging". Proceedings of the Conference on Empirical Methods in Natural Language Processing pp 133-142, University of Pennsylvania

No context found.

Ratnaparkhi, A.: 1996, `A Maximum Entropy Model for Part-of-Speech Tagging '. In: E. Brill and K. Church (eds.): Proceedings of the Conference on Empirical Methods in Natural Language Processing. pp. 133--142.

No context found.

A. Ratnaparkhi. 1996. A maximum entropy model for part-of-speech tagging. In Proc. EMNLP.

No context found.

Adwait Ratnaparkhi. A maximum entropy model for part-of-speech tagging. In Eric Brill and Kenneth Church, editors, Proceedings of the Conference on Empirical Methods in Natural Language Processing, pages 133--142. Association for Computational Linguistics, Somerset, New Jersey, 1996.

No context found.

Adwait Ratnaparkhi. 1996. A maximum entropy model for part-of-speech tagging. In EMNLP 1, pages 133--142.

No context found.

Adwait Ratnaparkhi. 1996. A maximum entropy model for part-of-speech tagging. In Proceedings of EMNLP 1996, pages 133--142.

No context found.

A. Ratnaparkhi. A maximum entropy model for part-of-speech tagging. In Proceedings of the First Conference on Empirical Methods in Natural Language Processing, pages 133--142, 1996.

No context found.

A. Ratnaparkhi. A maximum entropy model for part-of-speech tagging. In Proceedings of the Conference on Empirical Methods in Natural Language Processing, pages 133--142. Association for Computational Linguistics, Somerset, New Jersey, 1996.

No context found.

Ratnaparkhi, A. A maximum entropy model for part-of-speech tagging. In Conference on Empirical Methods in Natural Language Processing (May 1996), University of Pennsylvania, pp. 133-142.

No context found.

Adwait Ratnaparkhi. A maximum entropy model for part-of-speech tagging. In Eric Brill and Kenneth Church, editors, Proceedings of the Conference on Empirical Methods in Natural Language Processing, pages 133--142. Association for Computational Linguistics, Somerset, New Jersey, 1996.

No context found.

Ratnaparkhi, A. (1996) A maximum entropy model for part-of-speech tagging. In: Proceedings of the First Conference on Empirical Methods in Computational Linguistics (EMNLP 1996), 133-142.

No context found.

A. Ratnaparkhi. 1996. A Maximum Entropy Model for Part-of-Speech Tagging. In Proc. of the EMNLP'96.

No context found.

, Philadelphia, PA, USA, 1996.

No context found.

, Philadelphia, PA, USA, 1996.

No context found.

Adwait Ratnaparkhi. 1996. A maximum entropy model for part-of-speech tagging. In Proceedings of the Conference on Empirical Methods in Natural Language Processing EMNLP-96, Philadelphia, PA.

No context found.

Adwait Ratnaparkhi. 1996. A Maximum Entropy Model for Part-Of-Speech Tagging. Proceedings of Empirical Method for Natural Language Processings, pages 133--142.

First 50 documents

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