Learning Linguistically Valid Pronun
نویسندگان
چکیده
We describe an algorithm to learn word pronunciations from acoustic data. The algorithm jointly optimizes the pronunciation of a word using (a) the acoustic match of this pronunciation to the observed data, and (b) how “linguistically reasonable” the pronunciation is. Variations of word pronunciations in the recognition dictionary (which was created by linguists), are used to train a model of whether new hypothesized pronunciations are reasonable or not. The algorithm is well-suited for proper name pronunciation learning. Experiments on a corporate name dialing database show 40% error rate reduction with respect to a letter-to-phone pronunciation engine.
منابع مشابه
Learning linguistically valid pronunciations from acoustic data
We describe an algorithm to learn word pronunciations from acoustic data. The algorithm jointly optimizes the pronunciation of a word using (a) the acoustic match of this pronunciation to the observed data, and (b) how “linguistically reasonable” the pronunciation is. Variations of word pronunciations in the recognition dictionary (which was created by linguists), are used to train a model of w...
متن کاملChallenges of culturally and linguistically different healthcare students in learning environments
The increased number of international studentsin higher education systems is recognizedas beneficial not only economically but also interms of preparation of the workforce for theglobal environment. It is believed that diversityin the student cohort can also be beneficial fordomestic students in terms of increasing culturalawareness and achieving cultural competencygoals. Culturally and linguis...
متن کاملNamed Entity Transliteration Generation Leveraging Statistical Machine Translation Technology
Automatically identifying that different orthographic variants of names are referring to the same name is a significant challenge for processing natural language processing since they typically constitute the bulk of the out-of-vocabulary tokens. The problem is exacerbated when the name is foreign. In this paper we address the problem of generating valid orthographic variants for proper names, ...
متن کاملLinguistic and Non-Linguistic Influences on Learning Biases for Vowel Harmony
This paper addresses the question of the domain-specificity of learning biases for phonological processes. In two artificial grammar learning experiments we explore the role of learning biases in shaping the distribution of phonological patterns across the world’s languages. In Experiment 1, we demonstrate that learners are biased toward phonological patterns that occur in natural language, as ...
متن کاملSound symbolism facilitates early verb learning.
Some words are sound-symbolic in that they involve a non-arbitrary relationship between sound and meaning. Here, we report that 25-month-old children are sensitive to cross-linguistically valid sound-symbolic matches in the domain of action and that this sound symbolism facilitates verb learning in young children. We constructed a set of novel sound-symbolic verbs whose sounds were judged to ma...
متن کامل