Yannick Marchand and Robert I. Damper. A MultiStrategy Approach to Improving Pronunciation by Analogy. Computational Linguistics 26(2). 2000.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....to heavily constrain the search only entire words can be found and switching between words is disallowed as it would create new, and possibly, nonsense, words. Permitting switching could lead to other applications such as pronunciation generation (i.e. letter to sound) for novel words [81, 86] which could be used to address out of vocabulary (OOV) issues. The inverse task of recognizing novel words (i.e. sound to letter) would be another possibility. Analysis of sonorant sequences 3.1 Introduction Previous work by the author on unit selection [162] studied speech production and ....

Y. Marchand and R. I. Damper, "A multistrategy approach to improving pronunciation by analogy," Computational Linguistics, vol. 26, no. 2, pp. 195--219, June 2000.

....rules to the input word. PbA systems use some measure of similarity between words to retrieve partial pronunciations for fragments of the input word, which are then concatenated to obtain the final pronunciation. An excellent review (and critique) of these two approaches can be found in [3], which, to our knowledge, gives the only extension of the 1 Here, the term grapheme is used with the sense of functional spelling unit corresponding to a single phoneme (cf. 1] In principle, we allow the possibility that a grapheme may correspond to multiple phonemes in order to handle ....

....databases of spelling pronunciation pairs. Our work continues in this tradition. Specifically, we developed probabilistic techniques to align spellings with pronunciations automatically. From the resulting alignments a set of grapheme to phoneme (GP) correspondences (following the terminology in [3]) is induced. An n gram model trained on data aligned according to this set of correspondences can then be applied for bidirectional conversion between spelling and pronunciation. In the next section we present the theoretical formulation of our model. We devote most of the paper to describing ....

[Article contains additional citation context not shown here]

Marchand, Y., and R. I. Damper, "A multi-strategy approach to improving pronunciation by analogy". Computational Linguistics, 26(2):195-219, 2000.

....hypothesising a partial pronunciation for each matched substring from the phonological knowledge, and concatenating the partial pronunciations. The variant of PbA evaluated here is based on Dedina and Nusbaum s PRONOUNCE [4] but with several further enhancements as detailed by Marchand and Damper [5]. In PRONOUNCE, a data structure called the pronunciation lattice is built from matching substrings in the input word and the dictionary entries. This is a graph containing information about the position and total number of matched substrings, and their partial pronunciations. A possible ....

....and Nusbaum s partial matching. Second, multiple (five) heuristics are used to score the candidate pronunciations. Individual scores are then multiplied together to produce a final overall score. The best scoring pronunciation on this basis is then selected as output. Marchand and Damper show [5] that this multi strategy approach gives statistically significant performance improvements over simpler versions of PbA. Each word of the dictionary was processed in turn by removing it from the dictionary and assembling a pronunciation for it by analogy with the remainder. This can be seen as ....

Y. Marchand and R. I. Damper. "A multi-strategy approach to improving pronunciation by analogy". Manuscript submitted to Computational Linguistics , submitted.

....hypothesising a partial pronunciation for each matched substring from the phonological knowledge, and concatenating the partial pronunciations. The variant of PbA evaluated here is based on Dedina and Nusbaum s PRONOUNCE [4] but with several further enhancements as detailed by Marchand and Damper [5]. In PRONOUNCE, a data structure called the pronunciation lattice is built from matching substrings in the input word and the dictionary entries. This is a graph containing information about the position and total number of matched substrings, and their partial pronunciations. A possible ....

....and Nusbaum s partial matching. Second, multiple (five) heuristics are used to score the candidate pronunciations. Individual scores are then multiplied together to produce a final overall score. The best scoring pronunciation on this basis is then selected as output. Marchand and Damper show [5] that this multi strategy approach gives statistically significant performance improvements over simpler versions of PbA. Each word of the dictionary was processed in turn by removing it from the dictionary and assembling a pronunciation for it by analogy with the remainder. This can be seen as ....

Y. Marchand and R. I. Damper. "A multi-strategy approach to improving pronunciation by analogy". Manuscript submitted to Computational Linguistics.

No context found.

Yannick Marchand and Robert I. Damper. A MultiStrategy Approach to Improving Pronunciation by Analogy. Computational Linguistics 26(2). 2000.

No context found.

Marchand Y. and Damper R. I. 2000. "A multi-strategy approach to improving pronunciation by analogy," Computational Linguistics, 26(2), 195--219.

No context found.

Yannick Marchand and Robert I. Damper, "A multistrategy approach to improving pronunciation by analogy, " Computational Linguistics, vol. 26, no. 2, pp. 195--219, 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