We test both bottom-up and top-down approaches in learning the phonemic status of the sounds of English and Japanese. We used large corpora of spontaneous speech to provide the learner with an input that models both the linguistic properties and statistical regularities of each language. We found both approaches to help discriminate between allophonic and phonemic contrasts with a high degree o...

This paper addresses the problem of speaker-independent connected numeral recognition over telephone lines. Increasing the vocabulary from digits (0-9) to numerals (0-99) opens for more user-friendly services, but it also introduces many new, language-specific problems. This paper investigates morphological, phonemic and allophonic variations in the pronunciation of numerals in Norwegian. If im...

The distribution of the allophones of /?/in certain contexts involves free variation and gradient preferences. An organized survey was conducted to elicit the judgments of 37 native Persian speakers concerning the well-formedness of /?/allophonic behavior in five different phonological positions. The results showed that the differences in judgment between the various categories are not just t...

Phonological rules relate surface phonetic word forms to abstract underlying forms that are stored in the lexicon. Infants must thus acquire these rules in order to infer the abstract representation of words. We implement a statistical learning algorithm for the acquisition of one type of rule, namely allophony, which introduces context-sensitive phonetic variants of phonemes. This algorithm is...

This paper investigates the effects of allophones on the performance of Korean speech recognition systems. Along with a baseline phone-like unit (PLU) set consisting of phonemes, 31 allophone-based PLU sets are designed by systematically considering 5 major Korean allophonic constraints which can describe all the PLU sets currently used for Korean speech recognition systems. Experiments for pho...

Allophones are diverse phonetic instantiations of a single underlying sound category. As such, they pose a peculiar problem for infant language learners: These variants occur in the ambient language, but they are not used to encode lexical contrasts. Infants' sensitivity to sounds varying along allophonic dimensions declines by 11 months of age, suggesting that there must be information to phon...

This study was designed to identify English speech contrasts that might be appropriate for the computerbased auditory-visual training of Spanish learners of English. It examines auditory-visual and auditory consonant and vowel confusions by Spanish speaking students of English and a native English control group. 36 Spanish listeners were tested on their identification of 16 consonants and 9 vow...

We present a Wizard-of-Oz experiment examining phonetic accommodation of human interlocutors in the context human-computer interaction. Forty-two native speakers German engaged dynamic spoken interaction with simulated virtual tutor for learning language called Mirabella. Mirabella was controlled by experimenter and used either natural or hidden Markov model-based synthetic speech to communicat...

Tone merging and tonal confusion have been an interesting issue for a long time. This paper tries to investigate whether there is the phenomenon of merging tone in Hailu Hakka production. We focus on sandhi related tone pairs and level tone pairs. The results show that tone 24 and tone 33 is undergoing a sound change of merging in the younger speakers. It is proposed that both allophonic rules ...