Transdisciplinary & Innovation Grants awarded for 2018
The following projects have been awarded funding this round:
Project: Effective Digital Solutions for Sustaining Indigenous Languages
Steven Bird & Faith Baisden, Charles Darwin University & First Languages Australia
There is a critical need to revitalise and maintain Australia's Indigenous languages while there is still time. Digital technologies make it easy to record spoken language, connect dispersed speaker populations, and support language learning. This project will investigate three questions:
(a) what technologies are in regular use and how effective are they?
(b) what ideas do community members have for solutions that address unmet needs?
(c) what are the criteria for evaluating digital solutions?
Working closely with representatives from Indigenous language centres, we will prepare a Compendium of Ideas for Effective Digital Solutions for Sustaining Indigenous Languages.
Project: Category clustering: Testing the learnability advantage of a typologically preferred language structure
John Mansfield, University of Melbourne
Recent research has shown that in languages with complex word structures, there is a ‘category clustering’ bias to place elements with the same grammatical category in the same position in the word. Most languages conform to this bias, though some languages do not. We hypothesise that most languages have evolved this way because category-clustering is easier to learn and process. In order to test this hypothesis we have devised an artificial language experiment that requires participants to learn one of two invented ‘toy languages’ – one of which has category clustering, and the other of which does not.
Project: Coding actions: A cross-disciplinary approach to sign and gesture
Jennifer Green, University of Melbourne
The application is for funding to develop cross-disciplinary collaborations between linguists who work on sign languages, those who work on gesture, and those who have interests in both research fields. We seek funds for a workshop specifically designed to look at a defined set of kinesic-visual data, taking these varying disciplinary perspectives. The outcomes of this project will have benefits for the development of searchable corpora of sign languages and gesture, as well as having theoretical implications for some questions that remain unanswered and are of interest in these inter-connected fields of research.
Project: Learning language with a robot in the school classroom
Scott Heath, University of Queensland
This project is a collaboration between CoEDL engineers and CoEDL educators that aims to utilise a robot to support the teaching of language materials within a school classroom. While our previous work demonstrated how digital language resources can be adapted to a robot, further insights from educators, and integration with lesson plans are required for a robot to become an effective learning aid. Both these requirements are crucial to the educational outcomes from the robot, to the repeated use of the robot within a classroom, and to the robot playing a role in revitalisation of a language.
Project: Bininj Kunwok online university language course
Cathy Bow, Australian National University / Charles Darwin University
Currently in Australia there are few opportunities for University students to study Indigenous languages. The development of a Digital Language Shell is bridging that gap by enabling Indigenous language authorities to develop University level courses on their own terms. This project expands on the pilot program of 4 units of Bininj Kunwok delivered using the Shell in 2016 (see http://language-shell.cdu.edu.au/report-on-pilot-project/). It will develop a full semester course for delivery at undergraduate or postgraduate level, using an online platform, under the authority of the Bininj Kunwok Language Project Reference Group.
Project: Understanding dementia and ageing experiences of Indigenous Australians: An appreciative inquiry
Jacki Liddle, University of Queensland
Indigenous Australians experience dementia at five times the general population rates. It is a national health priority with issues of education and care provision highlighted. There is not a word for dementia within Indigenous Australian languages, and preferred ways of communication and support are not widely understood. To deliver optimal care, understanding how dementia is conceptualised and how to optimally communicate is essential. Yet Indigenous beliefs and practices around dementia have not been embedded within formal Aged Care. This study aims to explore how dementia and ageing are described and understood and what is working well for Indigenous Australians.
Project: Developing a web-based interface for Persephone, an automatic phoneme transcription tool, with Me'phaa Vátháá, Nafsan, Māori and Bininj Kunwok languages
Ben Foley, University of Queensland
This project brings together linguists and software developers to (1) iteratively design and develop a web interface for the phonemic transcription program, Persephone; and then (2) evaluate the interface for automatic phonemic transcription for Me'phaa Vátháá, Nafsan, Māori and Bininj Kunwok languages. Persephone's purpose is not to replace a linguist, but rather to produce a first pass phonetic transcription that the language worker can refine. It harnesses recent breakthroughs in machine learning to reduce manual data entry, freeing the linguist to focus on addressing linguistic questions. The web interface will provide access to Persephone for ordinary working linguists.
Project: Ultrasound tongue imaging in production and perception: implications for second language and second dialect acquisition
Chloe Diskin, University of Melbourne
Ultrasound tongue imaging uses general medical ultrasound technology to image movements of the tongue during speech. This research will draw on a large corpus of ultrasound recordings from Irish and Chinese migrants, and mainstream Australian English speakers, to find out whether prolonged exposure to Australian English results in adoption of more Australian-like pronunciation and articulation. This will be supported by a listening exercise, whereby people will listen to and make judgements about samples of recorded speech from the corpus. The study will help determine which articulatory and acoustic correlates are important for listeners when they process speech by L2 speakers.
Project: Tracking evolution in speech perception: A longitudinal intra-listener study
Debbie Loakes, University of Melbourne
This project offers a rare opportunity to analyse processing behaviour through analysis of longitudinal data. It is concerned with how consistent people's behaviour is over time when categorising vowels – in general, and crucially in the context of a sound change in progress.
The longitudinal data comprise responses to a listening test, from mainstream Australian listeners (2012, 2015 and through this project 2018) and Aboriginal English listeners (2016, now 2018). Longitudinal data such as this provides a unique opportunity to address the intersection of language processingandlanguageevolution,andtoassess whether any change over time is categorical or incremental.
Project: The role of technology in challenging word learning scenarios
Kristyn Hensby, University of Queensland
Technology has become pivotal in children’s education. However, technology has been cited as a poor educational tool in contrast to face-to-face interaction. Social robots boast features such as joint attention and heightened engagement, compared to other technologies, that may increase educational impact.
Our study will investigate whether children can learn to associate novel words and objectsbasedonmediumofinstruction. The medium will be either a robot with different levels of social ability, an iPad or a person. This project will bring together an interdisciplinary team of psychologists, linguists and roboticists from different nodes, programs and threads of CoEDL.
Project: Massive automated morpho-phonological analysis
Erich Round, University of Queensland
Linguists are nowadays assembling datasets covering the vocabularies of thousands of the world’s languages, helping us to quantify cross-linguistic variation and infer common ancestral lineages around the world. But words also often have internal structure, as in ‘black-bird’ or ‘fif-ty’, which typically is not explicitly indicated in such datasets but can also reveal historical connections and cross-linguistic patterns. We aim to build computational tools that can discover word structure within large, modern datasets, and build up overviews of how word pieces are assembled, both in terms of their component speech sounds and their meanings, across 270 indigenous languages of Australia.
Project: Forced-alignment protocols for minority languages
Simon Gonzalez, Australian National University
Forced-aligners have revolutionised sociophonetics. By automating the process of segmenting a time-aligned orthographic transcription into individual phonemes, they have vastly increased the size of datasets utilised in phonetic work, thereby enhancing the power and sensitivity of analyses conducted. Forced-aligners have mainly been employed for major languages, and there is a paucity of research exploring their utility for lesser-described languages. This project aims to assess the applicability and efficiency of existing forced-aligner technology on lesser-described languages, with the ultimate goals of creating guidelines and tools to assist researchers in preparing these languages for forced-alignment, facilitating and enhancing analyses thereof.
Project: Everyday noise: Do we learn without knowing? Identifying key brain networks present during natural language learning contexts
Kartik Iyer, University of Queensland
Understanding how the human brain responds to noise during language learning is critical to predicting brain behaviour during language processing and our capacity to optimise learning strategies. Although it is accepted that noise disrupts cognitive function in many situations, recent work demonstrates that white noise can actually enhance learning. Our study will combine EEG (monitoring electrical brain activity) with brain “connectomics” (computational methods) to challenge the prevailing view that noise is detrimental to spoken language comprehension and verbal learning. Combining EEG with emerging computational connectomics tools will provide deeper insights into how brain networks are shaped during language learning.
Project: Linguistic reconstruction in the age of Bayes
Robert Mailhammer, Western Sydney University
While quantitative methods have flourished for language subgrouping, proto-language reconstruction has lacked quantitative measures. We currently have no means of comparing proposed reconstructions objectively. This project will create a pilot system that can gauge the likelihood of a reconstruction given the sound changes it relies on, as a function of the likelihood of those sound changes. These likelihoods we will assess by their attested frequency in a catalogue of sound changes. Assessing the likelihood of both sound changes has significant implications for language acquisition because this is one key locus of language change.