School of Languages and Linguistics - Research Publications
Permanent URI for this collection
Search Results
1 - 10 of 33
-
ItemКрылатые выражения из советских кинофильмов как элементы национальной идентичности(Издательство "Научный консультант", 2019)The paper is dedicated to the questions of the reflection of national identity in the headlines of newspapers Komsomol'skaya Pravda – Moscow, Izvestiya and Literaturnaya Gazeta for the period from 1st January 2017 until 1st July 2018, drawing upon examples of winged words adapted from Soviet films. Specific winged phrases considered uniquely reflective of Russian culture are highlighted. An attempt is made to explain the main reasons behind winged phrases transformations. The paper stresses the value of learning winged phrases from Soviet films in practical classes of Russian taught as a foreign language.
-
ItemBuilding Speech Recognition Systems for Language Documentation: The CoEDL Endangered Language Pipeline and Inference System (ELPIS)(ISCA, 2018)Machine learning has revolutionized speech technologies for major world languages, but these technologies have generally not been available for the roughly 4,000 languages with populations of fewer than 10,000 speakers. This paper describes the development of ELPIS, a pipeline which language documentation workers with minimal computational experience can use to build their own speech recognition models, resulting in models being built for 16 languages from the Asia-Pacific region. ELPIS puts machine learning speech technologies within reach of people working with languages with scarce data, in a scalable way. This is impactful since it enables language communities to cross the digital divide, and speeds up language documentation. Complete automation of the process is not feasible for languages with small quantities of data and potentially large vocabularies. Hence our goal is not full automation, but rather to make a practical and effective workflow that integrates machine learning technologies.
-
ItemJapanese Vowel Devoicing Modulates Perceptual Epenthesis(Australian Speech Science and Technology Association, 2018)
-
ItemPhrasing and constituent boundaries in Lifou French(Australasian Speech Science and Technology Australia (ASSTA), 2019)
-
ItemPredictability, Word Frequency and Japanese Perceptual Epenthesis(Australasian Speech Science and Technology Association Inc., 2019)Speakers typically invest less effort in the articulation of sounds and words that are highly predictable from their contexts. Recent research reveals a perceptual corollary to this behaviour, showing that listeners pay less attention to acoustic signal in predictable contexts. The present paper expands on this finding by testing the acceptability and discriminability of sequences of speech with varying levels of predictability. Stimuli are contrast pairs and are either phonotactically attested or else contain an illicit nonhomorganic consonant cluster. Such clusters violate Japanese phonotactics and have been found to elicit perceptual epenthesis in Japanese listeners. The results show that unattested consonant clusters are perceived as more acceptable in high-frequency sequences than in low-frequency sequences.
-
ItemAcoustic correlates of lexical stress in Wubuy(Australasian Speech Science and Technology Association Inc., 2019)We examined the acoustic correlates of lexical stress in the non-Pama-Nyungan language Wubuy (Northern Territory, Australia). We tested two hypotheses about stress: that stress is determined by (1) a combination of syllable position in prosodic word and quantity sensitivity, or (2) by position alone. To test these hypotheses, we elicited trisyllabic noun roots differing in position of heavy syllables in frame-final environments from 3 speakers. We found that both position and predicted stress based on prior phonological descriptions could account for many correlates (segment and syllable duration, f0, intensity, vowel formants) although overall syllable position appeared to account for more of the variance.
-
ItemPhonetic exponence of word-level stress in Ashaninka(ARAWAK)(Australasian Speech Science and Technology Association Inc., 2019-08-02)This quantitative study examines the phonetic exponents of word-level stress in Ashaninka, an Arawak language of Peru, ISO-639-3 cni; glottocode asha1243. The variety under study is spoken in the Districts of Rio Negro, Satipo, Mazamari, and Llayla of the Satipo province, and District of Pichanaki of the Chanchamayo province of Junín Region. The analysis of Ashaninka word-level stress is based on the audio corpus of elicited speech made during focused fieldwork in the research community. The study results indicate that the right edge oriented primary stress is cued by two robust phonetic exponents, such as duration and intensity. The left edge oriented secondary stress is expressed via intensity. Vowel quality is not a statistically significant correlate to stress in the elicited data, except for the mid back /o/ whose formant structure is indicative of the levels of stress.
-
ItemCross-dialectal speech processing: perception of lexical stress by Indian English listeners(Australasian Speech Science and Technology Association Inc., 2019-08-09)In English, lexical stress provides essential information guiding lexical activation. However, little is known about the processing of lexical stress in postcolonial Englishes. The present study examines the perception of lexical stress in disyllabic words by adult speakers of Standard Indian English. Results show that in iambic words (second syllable stressed), participants perform at about 54% accuracy, regardless of social background. In trochaic words, participants with private schooling perform significantly better (60% accuracy; p<0.05) than those with a government school background, approaching the level of accuracy reported for Australian English listeners. Our results suggest that processing of the commonly occurring trochaic condition is easier for participants from private schools, while processing of the rare iambic pattern is not eased by such experience. L1 background and onset of learning English show no systematic effect on participants’ performance. Variability in Standard Indian English is shaped mainly by schooling and not L1 background.
-
ItemSecond language fluency: re-thinking utterance fluency from a phonetics-phonology interface(Australasian Speech Science and Technology Association Inc., 2019-08-09)Research into second language fluency has called for cross-linguistic studies to rule out measures that can be attributed to intra-speaker variation. However, cross-linguistic comparisons in fluency studies are problematic. Research in the area has not necessarily taken into consideration differences across languages such as syllable structure, phonotactics, durational cues to prominence and prosodic levels, and idiosyncratic nature of pause duration. The preliminary results of this study into L2 fluency in Chilean Spanish speakers of English revealed that speed and pause phenomena were mostly idiosyncratic, and that segments rather than syllables could be a more reliable measure. Durational cues for phrasal level prominence were not implemented consistently in the L1 and preboundary lengthening in the L2 was not necessarily being used to signal prosodic constituent boundaries. It may be useful to re-operationalize measures used in L2 fluency studies from a phonetics-phonology interface perspective
-
ItemTense-lax contrasts in Indian English vowels: transfer effects from L1 Telugu at the phonetics-phonology interface(Australasian Speech Science and Technology Association Inc., 2019-08-09)We investigate the effects of L1 Telugu on tense-lax contrasts in Indian English vowels. While English has a tense-lax contrast in high vowels, / iː, ɪ, uː, ʊ/, with duration as an additional cue, Telugu has only a shortlong contrast, /i, iː, u, uː/, though these also have the lax allophones /ɪ, ɪː, ʊ, ʊː/ as a result of vowel harmony (VH), triggered by a following low vowel. We examine whether L1 transfer effects are limited to the ‘base’ phonological inventory (e.g. ‘borrowing’ the Telugu length contrast for English), or whether speakers access the spectrally closer VH allophones from Telugu. The results reveal something more complex, with some speakers showing tense-lax allophones also for Telugu length contrasts. In L1-L2 transfer, these speakers exapt these phonetically laxer short allophones for the English lax vowels. The other speakers, showing less tense-lax variation all round in L1, create entirely new phonetic categories for the English lax vowels.