Escudero, G., Marquez, L. & Rigau, G., Boosting applied to word sense disambiguation. Proceedings of ECML-00, 11th European Conference on Machine Learning, eds. R.L.D. Mantaras & E. Plaza, Springer Verlag, Heidelberg, DE: Barcelona, ES, pp. 129--141, 2000. Published in the "Lecture Notes in Computer Science" series, number 1810.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....[196] and text filtering [198] where an adapted version of the popular AdaBoost algorithm [82, 83, 197] is presented for information retrieval tasks. Other applications of AdaBoost variants to NLP tasks include: PoS tagging [2] PP attachment disambiguation [2] word sense disambiguation [72, 74], and full parsing [93] 13] is another relevant example of the combination of classifiers applied to information retrieval, in which the combination of classifiers allows the use of a big set of unlabelled examples (semi supervised approach) to iteratively improve the classification accuracy in ....

.... 150, 112] 67] 203, 166] Text categorization and filtering [117, 69, 230] 117, 190, 119, 142, 196] 162, 163] Dialogue act tagging [192] 193, 192] Co reference and anaphora resolution [5, 143] Cue phrase identification [122] IBL NNs EC SVM Clust WSD [159, 157, 84, 73] 150, 224] [182, 72, 74, 165] [202] Text categorization and filtering [183, 238, 237, 239] 233] 198, 196, 13] 98, 69, 99] Co reference and anaphora resol. 39] 148, 147] 35, 34] Rocchio RI ILP LSM GAs ME WSD [74] Text categorization and filtering [185, 91, 118, 196, 64] 49, 51, 50, 53, 134, 213] 51, ....

[Article contains additional citation context not shown here]

G. Escudero, L. M`arquez, and G. Rigau. Boosting Applied to Word Sense Disambiguation. In To appear in Proceedings of the 12th European Conference on Machine Learning, ECML, Barcelona, Spain, 2000.

....skewed words, where we can expect better performance than average. 4.8 Overall DSO: state of the art results In order to compare decision lists with other state of the art algorithms we tagged all 191 words in the DSO corpus. The results in (Ng, 1997) only tag two subsets of all the data, but (Escudero et al. 2000a) implement both Ng s example based (EB) approach and a Naive Bayes (NB) system and test it on all 191 words. The same test set is also used in (Escudero et al. 2000b) which presents a boosting approach to word sense disambiguation. The features they use are similar to ours, but not exactly. The ....

....of the art algorithms we tagged all 191 words in the DSO corpus. The results in (Ng, 1997) only tag two subsets of all the data, but (Escudero et al. 2000a) implement both Ng s example based (EB) approach and a Naive Bayes (NB) system and test it on all 191 words. The same test set is also used in (Escudero et al. 2000b) which presents a boosting approach to word sense disambiguation. The features they use are similar to ours, but not exactly. The precision obtained, summarized on Table 7 show that decision lists provide state ofthe art performance. Decision list attained 0.99 coverage. 5 Cross tagging: hand ....

Escudero, G., L. Mrquez and G. Rigau. Boosting Applied to Word Sense Disambiguation. Proceedings of the 12th European Conference on Machine Learning, ECML 2000. Barcelona, Spain. 2000.

.... is familiar with the related concepts (see [12] otherwise) It has to be noted that this algorithm has been applied, with significant success, to a number of NLP disambiguation tasks, such as: Part of speech tagging and PP attachment [2] text categorization [13] and word sense disambiguation [6]. The purpose of boosting is to find a highly accurate classification rule by combining many weak classifiers (or weak hypotheses) each of which may be only moderately accurate. The weak hypotheses are learned sequentially, one at a time, and, conceptually, at each iteration the weak hypothesis ....

Escudero, G; M`arquez, L. and Rigau, G.: Boosting Applied to Word Sense Disambiguation. In Proceedings of the 12th European Conference on Machine Learning, ECML, Barcelona, Spain, 2000.

.... Many standard ML algorithms for supervised learning have been applied, such as: Decision Lists (Yarowsky, 1994; Agirre and Martinez, 2000) Neural Networks (Towell and Voorhees, 1998) Bayesian learning (Bruce and Wiebe, 1999) Exemplar Based learning (Ng, 1997a; Fujii et al. 1998) Boosting (Escudero et al. 2000a) etc. Unfortunately, there have been very few direct comparisons between alternative methods for WSD. In general, supervised learning presumes that the training examples are somehow re ective of the task that will be performed by the trainee on other data. Consequently, the performance of ....

....example. In the experiments explained in section 4, the EB algorithm is run several times using di erent number of nearest neighbours (1, 3, 3 Although the use of MVDM metric (Cost and Salzberg, 1993) could lead to better results, current implementations have prohivitive computational overheads (Escudero et al. 2000b) 5, 7, 10, 15, 20 and 25) and the results corresponding to the best choice are reported 4 . Exemplar based learning is said to be the best option for WSD (Ng, 1997a) Other authors (Daelemans et al. 1999) point out that exemplar based methods tend to be superior in language learning ....

[Article contains additional citation context not shown here]

G. Escudero, L. Marquez, and G. Rigau. 2000a. Boosting Applied to Word Sense Disambiguation. In Proceedings of the 12th European Conference on Machine Learning, ECML, Barcelona, Spain.

....decreases slightly and monotonically, as it approaches the maximum number of rules reported 4 . According to the plot in figure 2, no overfitting is observed while increasing the number of rules per word. Although it seems that the best strategy could be learn as many rules as possible , in [7] it is shown that the number of rounds must be determined individually for each word since they have different behaviours. The adjustment of the number of rounds can be done by cross validation on the training set, as suggested in [1] However, in our case, this cross validation inside the ....

....on supervised WSD. In addition, a faster variant has been suggested and tested, which is called LazyBoosting. This variant allows the scaling of the algorithm to broad coverage real WSD domains, and is as accurate as AdaBoost.MH. Further details can be found in an extended version of this paper [7]. Future work is planned to be done in the following directions: ffl Extensively evaluate AdaBoost.MH on the WSD task. This would include taking into account additional attributes, and testing the algorithms in other manually annotated corpora, and especially on sense tagged corpora ....

Escudero, G., M`arquez, L. and Rigau, G. Boosting Applied to Word Sense Disambiguation. Technical Report LSI-00-3-R, LSI Department, UPC, 2000.

....pseudo words. Many standard ML algorithms for supervised learning have been applied to WSD, including: Decision Lists (Yarowsky, 1994) Neural Networks (Towell and Voorhees, 1998) Bayesian learning (Bruce and Wiebe, 1999) Exemplar based learning (Ng, 1997a; Fujii et al. 1998) and Boosting (Escudero et al. 2000a) Further, in (Mooney, 1996) some of the previously cited methods are compared, jointly with Decision Trees and Rule Induction algorithms, on a very restricted domain. The performance of supervised ML based systems is usually calculated by testing the algorithm on a separate part of the set ....

....obtain from Internet arbitrarily large samples of word senses (Leacock et al. 1998; Mihalcea and Moldovan, 1999) 3. The use of unsupervised EM like algorithms for estimating the statistical model parameters (Pedersen and Bruce, 1998) 4. The application of efficient and accurate ML algorithms (Escudero et al. 2000a) and attribute representations (Escudero et al. 2000b) in order to deal with real size WSD problems. Solving the problem of knowledge acquisition is crucial for making the unsupervised WSD approach viable. It is our belief that the referred work, and especially second and fourth lines, ....

[Article contains additional citation context not shown here]

G. Escudero, L. M`arquez, and G. Rigau. 2000a. Boosting Applied to Word Sense Disambiguation. In To appear in Proceedings of the 12th European Conference on Machine Learning, ECML, Barcelona, Spain.

No context found.

Escudero, G., Marquez, L. & Rigau, G., Boosting applied to word sense disambiguation. Proceedings of ECML-00, 11th European Conference on Machine Learning, eds. R.L.D. Mantaras & E. Plaza, Springer Verlag, Heidelberg, DE: Barcelona, ES, pp. 129--141, 2000. Published in the "Lecture Notes in Computer Science" series, number 1810.

No context found.

Gerard Escudero, Lluis Marquez, and German Rigau. Boosting applied to word sense disambiguation. In European Conference on Machine Learning, pages 129-- 141, 2000.

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