Dan Melamed. 1997. A word-to-word model of translational equivalence. In Proceedings of ACL.  Home/Search   Document Details and Download   Summary   Related Articles   Check

.... Tourist Task corpus, which was generated in a semi automated way. The best results achieved were a word error rate of 9.3 and a sentence error rate of 44.4 . 1. INTRODUCTION The statistical approach is an adequate framework for introducing automatic learning techniques in Machine Translation [3, 14, 5, 15]. Under this framework, given an input string s from S (S is a finite input alphabet and S is the set of finite length strings over S) the probabilistic translation of s is an output string, e 2 E (E is a finite output alphabet) such that Pr(ejs) 1) Using Bayes theorem, and ....

I. Dan Melamed. "A Word-to-Word Model of Translational Equivalence". In Procs. of the ACL97. pp 490--497. Madrid Spain, 1997.

....in a semi automatic way. The best results obtained were a word error rate of 0.52 and a sentence error rate of 3.2, using the iterative search algorithm. 1 Introduction The statistical approach is an adequate framework for introducing automatic learning techniques in Machine Translation [2, 9, 14, 15]. Under this framework, given an input string s from S (S is a nite input alphabet) the probabilistic translation of s is an output string, e 2 E (E is a nite output alphabet) such that Pr(ejs) 1) Using Bayes theorem, and taking into account that Pr(s) is not a function of e, ....

I. D. Melamed. 1997. A Word-to-Word Model of Translational Equivalence. In Proceedings of the ACL97. pp 490-497. Madrid Spain.

....so that longrange dependencies and embedded structures can help the classification decision, as in language modeling [Martin et al. 99, Sawaf et al. 00] 2. 2 Document Clustering For document clustering we use a criterion based on the mutual information criterion as described in [Melamed 97] and other publications. The criterion is defined as follows: # # # # 2 N i i i n N c n w n n c w n c w n c w MI 1 1 ) log ) Here, c w n i is the count of a unit i w in the class c , n w n and ) c n the count of a unit i w and of a ....

I. Dan Melamed. A word-to-word model of translational equivalence. In 35 th Conference of the Association for Computational Linguistics (ACL'97), pp. 490-497, Madrid, 1997.

....measure. Anchor matching is used as the only measure for matching and the resolving of matching collision . It is interesting to note that our matching collision problem in file matching is comparable to the problem of indirect association in the study of word concurrence reported in Melamed (1997). His solution to indirect association , i.e. the competitive linking algorithm , is also similar to our solution, but looks much simpler than ours. As he pointed out: The simplicity of the competitive linking algorithm depends on the one to one assumption: each word translates to at most one ....

Melamed, I. Dan: 1997, `A word-to-Word Model of Translational Equivalence', In the proceedings of ACL97.

....with 99 precision and 46 recall when trained on 13 million words of the Hansard corpus, where recall was measured as the fraction of words from the bitext that were assigned some translation. Using the same model but less data, a French English software manual of 400,000 words, Resnik and Melamed (1997) reported 94 precision with 30 recall. While these figures are indeed impressive, more telling figures can only be obtained by measuring the effect of the alignment system on some specific task. Dagan and Church (1994) reports that their Termight system helped double the speed at which terminology lists could be compiled at ....

Melamed, I. D. (1997a) "A Word-to-Word Model of Translational Equivalence." Proceedings of the 35th Conference of the Association for Computational Linguistics, Madrid.

....equivalence on the word and phrase level. As Swedish is highly in ectional, a pure string based alignment approach would fail to nd many of the accurate alignments, we designed a simple pattern matcher to handle in ectional variants for the Swedish texts. Furthermore, along the lines of Melamed [Mel97] we iterate processing so that source and target candidates that become linked, are removed from the list of possible candidates, thereby shrinking the search space for alignments that remain to be done. After a given set of iterations it is possible to switch to function words and try to align ....

I. Dan Melamed. A word-to-word model of translational equivalence. In 35th Conference of the Association for Computational Linguistics (ACL'97), pages 490-497, Madrid, 1997.

....the approach seem uneconomic and presumably not very eOEcient and portable. However, this approach achieves high precision rates on small test corpora with low averages of 3.3 alternative translations and high averages of 1.4 correct translations. Another important approach to SD is the work of Melamed (1997). Melamed reports 42 precision and 35 40 recall using a gold standard for single best word to word translations (French English) The test corpus contains several thousand items and is a part of the training corpus. Melamed s best model can be viewed as a robust, economic, domain specic, and ....

Melamed, I. D. (1997). Word-to-word models of translational equivalence. Technical Report IRCS, Nr. 98-08, University of Pennsylvania.

....resource bottleneck, namely the lack of parallel corpora, in particular for languages that are less likely to be available in an electronic form. It would be interesting to compare the results of our model once we have results cross linguistically to models of word alignment (Brown et al. 1991; Melamed, 1997). These models get leverage from sentence alignment, which is the reason there is a reliance on parallel corpora, accordingly, using heuristics within the sentence to arrive at word level mapping. Our approach should, in principle, be able to do this mapping with no need for the overhead of ....

Melamed, I. Dan (1997). Word-to-Word Models of Translational Equivalence. Proceedings of the 35 th Annual Conference of the Association for Computational Linguistics.

....the approach seem uneconomic and presumably not very eOEcient and portable. However, this approach achieves high precision rates on small test corpora with low averages of 3.3 alternative translations and high averages of 1.4 correct translations. Another important approach to SD is the work of Melamed (1997). Melamed reports 42 precision and 35 40 recall using a gold standard for single best word to word translations (French English) The test corpus contains several thousand items and is a part of the training corpus. Melamed s best model can be viewed as a robust, economic, domain specic, and ....

Melamed, I. D. (1997). Word-to-word models of translational equivalence. Technical Report IRCS, Nr. 98-08, University of Pennsylvania.

....word alignment depends on the use of many sources of information in concert 1 . Distributional parallelism, coocurrence, string similarity (both between and within languages) and part of speech are some such information sources used in previous research (see, e.g. Tiedemann 1998, 1999a, 1999b; Melamed 1995, 1998). In the ETAP project we have so far concentrated on linguistically rich information sources, such as word similarity The research reported here was carried out within the ETAP project, supported by the Bank of Sweden Tercentenary Foundation as part of the research programme Translation and ....

Melamed, I. Dan 1998. Word-to-word models of translational equivalence. IRCS Technical Report #98--08. Department of Computer and Information Science, University of Pennsylvania.

....is called for. This approach allows us to achieve a broader coverage of potential correspondences across languages. 2. 2 Word to word alignment Recent research has demonstrated that statistical alignment models can be highly successful at extracting word correspondences from parallel corpora [3, 18]. These algorithms use global word cooccurrence data to derive links between words at the local sentence level. Our system uses a probabilistic translation model developed by Hiemstra [13] In this algorithm, each sentence pair is represented by a matrix. The rows represent source language words, ....

....Estimating expected counts via the Iterative Proportional Fitting Algorithm Once the Hiemstra algorithm has converged, we can approximate the most likely alignment on a sentence by sentence basis using the simple greedy algorithm given below. It is nearly equivalent to the one suggested by Melamed [18], but is applied to each sentence individually. 1) Define score: S jk = P (s j t k ) P (t k s j ) P (o jk ) 2) Sort pairs in descending order of their score (3) Take largest S jk and align s j and t k (4) Remove S j: and S :k from the score pool (5) Repeat steps (3) and (4) until all ....

[Article contains additional citation context not shown here]

I. Dan Melamed. Word-to-Word Models of Translational Equivalence. Technical Report IRCS Technical Report #98-08, University of Pennsylvania, 1998. Available at: http://www.cis.upenn.edu/~melamed.

No context found.

Dan Melamed. 1997. A word-to-word model of translational equivalence. In Proceedings of ACL.

No context found.

Dan Melamed, "A Word-to-Word Model of Translational Equivalence," Proceedings of 35th Conference of the Association for Computational Linguistics (ACL'97), Madrid, Spain, 1997.

No context found.

Dan Melamed. A word-to-word model of translational equivalence. In 35th Conference of the Association of Computational Linguistics (ACL'97), Madrid, Spain, 1997.

No context found.

I. Dan Melamed. 1998. Word-to-Word Models of Translational Equivalence. Technical Report 98-08, Dept. of Computer and Information Science, University of Pennsylvania, Philadelphia, USA.

No context found.

I. Dan Melamed. 1998. Word-to-Word Models of Translational Equivalence. Technical Report 98-08, Dept. of Computer and Information Science, University of Pennsylvania, Philadelphia, USA.

No context found.

I. Dan Melamed. 1998. Word-to-Word Models of Translational Equivalence. Technical Report 98-08, Dept. of Computer and Information Science, University of Pennsylvania, Philadelphia, USA.

No context found.

I. Dan Melamed. 1998. Word-to-Word Models of Translational Equivalence. Technical Report 98-08, Dept. of Computer and Information Science, University of Pennsylvania, Philadelphia, USA.

No context found.

I. Dan Melamed. 1998. Word-to-Word Models of Translational Equivalence. Technical Report 98-08, Dept. of Computer and Information Science, University of Pennsylvania, Philadelphia, USA.

No context found.

I. Dan Melamed. 1998. Word-to-Word Models of Translational Equivalence. Technical Report 98-08, Dept. of Computer and Information Science, University of Pennsylvania, Philadelphia, USA.

No context found.

Melamed, I. D. (1997). Word-to-word models of translational equivalence. Technical Report IRCS, Nr. 98-08, University of Pennsylvania.

No context found.

Melamed, I. D. (1997). Word-to-word models of translational equivalence. Technical Report IRCS, Nr. 98-08, University of Pennsylvania.

No context found.

Melamed, I. D. (1997). Word-to-word models of translational equivalence. Technical Report IRCS, Nr. 98-08, University of Pennsylvania.

No context found.

Melamed, I. Dan: 1998b, `Word-to-Word Models of Translational Equivalence', ICRS Technical Report 98-08, University of Pennsylvania.

No context found.

Melamed, I. D. (1997a) A Word-to-Word Model of Translational Equivalence. In Proceedings of ACL-97, Madrid.

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