Papers by Briony Williams
Annual Conference of the International Speech Communication Association, 2005
This paper reports on efforts to set up a research capability in speech technology for two minori... more This paper reports on efforts to set up a research capability in speech technology for two minority languages (Welsh and Irish), where the focus is on ensuring that this capability will outlive the project that provided the initial impetus (the WISPR project). The existing situation at the start of the project is summarized, together with the challenges (setting up plant and equipment from scratch, training researchers in specialised skills). Some innovative strategies are examined: remote working by researchers with regular intensive on-site work weeks, and the use of Internet technology as a medium for collaborative work patterns. It is hoped that these experiences may be useful to researchers in other minority languages who wish to set up a similar research capability.
Empirical Approaches to Language Typology, 1999
A speech database for Welsh was recorded in a studio from read text by a few speakers. The purpos... more A speech database for Welsh was recorded in a studio from read text by a few speakers. The purpose is to investigate the acoustic characteristics of Welsh speech sounds and prosody. It can also serve as a resource for future work in speech synthesis and recognition.
To begin with, the results of measurements on vowels and consonants in Welsh read speech indicate... more To begin with, the results of measurements on vowels and consonants in Welsh read speech indicate that stress in Welsh is determined by rhythm rather than inherent acoustic cues (of F0, duration and intensity). Further measurements of 'feet' (interstress intervals) support ...
We report a set of labelling criteria which have been developed to label prosodic events in clear... more We report a set of labelling criteria which have been developed to label prosodic events in clear, continuous speech, and propose a scheme whereby this information can be transcribed in a machine readable format. We have chosen to annotate prosody in a syllabic domain which is synchronised with a phonemic segmentation. A procedural de nition of syllables based on the grouping of phones is presented. The criteria for hand labelling the prominence of each syllable, tone-unit boundaries and the pitch movement associated with each accented syllable, are described. Work to automate this process is presented and experimental results evaluating its performance are included.
Interspeech, 1995
... Converting to North Welsh would require the following steps: [8] MJ Ball & GE Jones (eds.... more ... Converting to North Welsh would require the following steps: [8] MJ Ball & GE Jones (eds.) (1984) Welsh Phonology. Cardiff: University of Wales Press. [9] PA Taylor & SD Isard (1991) "Automatic diphone segmentation", EUROSPEECH, 1991, pp. 709-711. ...
Isochrony has been considered only in terms of stressed syllables. However, it may also be a rand... more Isochrony has been considered only in terms of stressed syllables. However, it may also be a random property of unstressed syllables, and a control experiment was deemed necessary. A handtranscribed database of 98 sentences, each produced by three speakers, formed the input to an algorithm calculating durations of feet, number of syllables per foot, and mean syllable duration within each foot. In each output dataset, feet were based on one of the following criteria: stressed, tense, unreduced, random, or arbitrary syllables (the latter based on ordinal numbers of syllables within the utterance). Calculations were made of the correlations between foot duration and number of syllables per foot, and between mean syllable duration and number of syllables per foot. The 'foot compression effect' was shown to be nonrandom, and due to linguistic rather than arbitrary factors. A detailed examination of actual syllable durations was then carried out. The main determinants of syllable duration were the number of constituent segments, and syllable status in terms of target/nontarget. A small but significant syllable shortening effect was also found, dependent on the number of syllables in the foot, which was linguistic rather than random.
Journal of Phonetics
An experiment was conducted to assess the relative contribution of Fâ‚€ pattern and segmental durat... more An experiment was conducted to assess the relative contribution of Fâ‚€ pattern and segmental duration to the perception of stress in Welsh. A stress-related minimal pair of words was recorded and edited to give stimulus continua and was resynthesized with 3 types of pitch pattern superimposed. 10 Welsh listeners made stress judgments on the randomized stimuli. Results show a complex interaction of Fâ‚€, segmental duration, and stress perception in Welsh. Pitch-prominence alone was no cue to stress, but functioned as such in terms of recognized intonational categories. A certain type of segmental durational effect also influenced stress judgments. The patterns of co-occurrence between pitch patterns and stress location that are found in the intonational units of the language were the means by which stress was indirectly perceived when the stimuli showed a change in Fâ‚€. (PsycINFO Database Record (c) 2012 APA, all rights reserved)
Introduction Segmentation is the division of a speech file into non-overlapping sections correspo... more Introduction Segmentation is the division of a speech file into non-overlapping sections corresponding to physical or linguistic units. Labelling is the assignment of physical or linguistic labels to these units. Both segmentation and labelling form a major part of current work in linguistic databases. 1.1.1 Segmental transcription The term `transcription' may be used to refer to the representation of a text or an utterance as a string of symbols, without any linkage to the acoustic representation of the utterance. This was the pattern followed by speech and text corpus work during the 1980's, such as the prosodically-transcribed Spoken English Corpus (Knowles et al. 1995). These corpora did not link the symbolic representation with the physical acoustic waveform, and hence were not fully machine-readable. A recent project, MARSEC (Roach et al. 1993), has generated these links for the Spoken English Corpus such that it is now a
A Welsh speech database intended for use in phonetic research requires careful annotation at seve... more A Welsh speech database intended for use in phonetic research requires careful annotation at several linguistic levels. The initial stage is that of labelling at the acoustic phonetic level, where the closure, burst and aspiration phases of a stop consonant are all separately labelled. The next is the phonemic stage, which can be derived from the former in most cases. Next is the syllabic stage, where each syllable is labelled in terms of its word status and in terms of lexical stress. The final stage is the lexical stage, where each word is labelled according to its word class. A statistical package can then be run over this data to yield information on the acoustic characteristics of Welsh speech sounds, and also about the nature of lexical stress in Welsh. In addition, it is hoped to derive rules for intonation patterns for use in an existing Welsh text-to-speech synthesiser.
Computational Linguistics
INTRODUCTION The Welsh language is one of the lesser-used and lesser-researched languages of Euro... more INTRODUCTION The Welsh language is one of the lesser-used and lesser-researched languages of Europe. This work represents the first known attempt at developing a speech synthesiser for Welsh. Because comparatively little is known about the acoustic characteristics of Welsh speech sounds, it was decided to use diphone concatenation rather than rule-based parametric synthesis. The software of an existing text-to-speech synthesis system for English (described in [1]) was adapted for use with Welsh. This software uses the PSOLA synthesis technique, as descibed in [2], [3]. The software can run on a SUN workstation or on a PC with an LSI DSP board. The number of Welsh phonemes included was 51, including 3 used only in English loanwords (/z/, affricates /ch/ and /jh/) and 3 used in restricted contexts (labialised /lw, nw, rw/). In total, there were 32 consonants and 19 vowels. Also, it was decided that the synthesiser should be able to handle English as well, due to the number of En
Literary and Linguistic Computing, 2000
The Journal of the Acoustical Society of America, 1986
Uploads
Papers by Briony Williams