PWorkshop Archives: Semester 1, 2005–2006

10 May 2005

Hiroshi Shimodaira

As mobile and wearable computing devices have become popular, a number of text input interfaces as substitutes for keyboard interface have been developed. If we assume much smaller devices than PDAs, i.e. cellular phones or a small touch-pad attached to human bodies for wearable computing, handwriting interface would be the most natural interface widely accepted by the users, while button interface is preferred by a limited number of users.

To take full advantage of handwriting interface, we assume that the users are allowed to write characters continuously without watching them. Under the assumed condition, characters are written one after another without pauses on a small writing area, with the result that written characters are heavily distorted and overlaid each other. Conventional handwriting recognition engines, however, can not handle this sort of handwritings. To tackle the problem, we employ the state-of-the-art automatic speech recognition technologies, and develop special handwriting input devices for different styles of wearable computing.

In the talk, I will show a prototype system and several handwriting input devices, one of which is for handwriting in the air. In addition to that, I will introduce a handwriting interface for the visually impaired people, which is one of the applications of the eyesfree handwriting interface.

27 Sep 2005

Jan-Peter de Ruiter (MPI Psycholinguistics, Nijmegen)

Listeners in a conversation do not only have to determine what they are going to say next, but also when to say it. From research in Conversation Analysis and my own quantitative work on conversation it is clear that listeners have an uncanny ability to anticipate the moment the current speaker will have finished their turn. This allows them to start their own contribution with hardly any temporal gap or overlap between two succesive turns. We have performed classical button press experiments involving fragments from natural conversations to determine which cues listener's use to anticipate, or 'project' the end of a speaker's turn. Specifically, we have focussed on two sources of information: a) lexico-syntax, and b) intonation. Perhaps surprisingly, we have found that having access to the lexico-syntax is both sufficient and necessary for accurate end-of-turn projection, while intonation is neither necessary nor sufficient.

04 Oct 2005

Doris Muecke (IfL Phonetics Cologne)

In the area of tonal alignment research, it has been shown that F0 targets are phonetically synchronised with acoustically determined segment boundaries in a fairly robust way. These boundaries are generally taken to be the boundaries of segments belonging to or relating in some way to the accented syllable, which has a phonological assocation with the tone or combination of tones corresponding to the F0 target(s). My talk takes up problems discussed by Bob Ladd in an earlier P-Workshop concerning the syllable-based hypothesis of tonal alignment in prenuclear and nuclear rises in Dutch. While Ladd et al. 2000 found a clear influence of syllable structure on the alignment of H targets in prenuclear position (caused by phonological vowel length), their findings of H alignment in nuclear position did not confirm this strong effect of syllable structure (Schepman, Mennen, Ladd in press).

The talk presents data from two experiments which are comparable to the Dutch study. However, the language is German and the search for possible anchors for F0 targets extends into the domain of articulatory gestures (using kinematic data from EMMA recordings). In the first experiment, we manipulated syllable structure by varying phonological vowel length, and in the second experiment, we manipulated accent position (nuclear vs. prenuclear).

We argue (i) that the syllable based-anchor hypothesis for tonal alignment might be too restrictive, in that not only the accented syllable but also adjacent syllables can serve to define anchors, and (ii) that F0 movements are synchronised with kinematic movements that can be modelled phonologically by sonority profiles (of accented and postaccented syllables).

11 Oct 2005

Peter Ladefoged (UCLA)

Linguistic classifications of speech sounds are apt to be biased by taxonomic traditions. such as those of the IPA, Jakobsonian distinctive features, and Chomsky and Halle phonological features. There are several reasons why none of these systems provide a principled basis for describing phonological systems.

1. Universal classificatory systems need to take into account when a sound in one language can be considered to be potentially distinct from a similar sound in another language.
2. Patterns of sounds in languages have a variety of different origins.
3. Some speech sounds do not fit into any pattern.

This paper will propose a theory of phonetics, a system for classifying the linguistic properties of speech sounds that takes these points into account - and leaves other problems for future work.

18 Oct 2005

Leoma Gilley (SIL)

15 Nov 2005

Olga Gordeeva (QMUC)

This study explores the possibility of an additional phonetic dimension to the English tense/lax vowel contrast involving differentiated vocal effort at the laryngeal level. The laryngeal dimension is assessed acoustically (Sluijter, 1995; Jessen 2002) and more directly by using electroglottography in conjunction with Marashek's (1997) model of the EGG signal. The EGG results are based on the MOCHA (Wrench, 2000) corpus.

22 Nov 2005

Ioulia Grichkovtsova (QMUC)

The main objective of this research is to investigate the production of affective speech by bilingual and monolingual children cross-linguistically. Recent cross-linguistic studies show that there is a number of available means to express emotions in speech (pitch range, rhythm, voice quality, etc), but their usage, level of importance and meaning vary in different languages. Crosscultural and cross-linguistic differences in affective speech may lead bilingual children to perceive and to express emotions differently in their two different languages. Given that bilingual children produce two languages that differ in some acoustic correlates for affective speech, it is likely that they may differ in their production of affective speech from the monolinguals in each of their two languages. This phenomenon has been widely attested in bilingual studies of other phonetic aspects.

A cross-linguistically comparable corpus of 6 bilingual Scottish-French children and 12 monolingual peers, aged between 7 and 10, was recorded according to the developed methodology. This talk presents results on pitch range, peak alignment and speech rate for four bilingual children and their eight monolingual peers, comparing their emotions and languages. The results show that bilingual children differentiate some emotions across their languages; happiness is the only emotion, which is differentiated by all the bilingual children. The majority of children (both bilingual and monolingual) realize differences between some emotions in each of the measurements, taken in this study. Monolingual children use analysed acoustic parameters in a much more homogeneous way than bilinguals. Some results of bilingual children do not strictly correspond to those of monolinguals. Having a wider range of means and ways for affective realizations, bilinguals may represent a particular group of speakers who express vocal emotions in a different manner than monolinguals.

13 Dec 2005

Bob Ladd and Dan Dediu

A correlation between two new mutations of brain growth and development-related genes and tonality was detected. It cannot be explained by geographic proximity, nor by a simple founder effect and we conjecture that it might represent the first case of causal link between variation in genes and linguistic features. This offers new perspectives on language evolution and the human linguistic capacity.