Alveo, a Human Communication Science Virtual Laboratory

Alveo, a Human Communication Science Virtual Laboratory Dominique Estival U. of Western Sydney d.estival@uws.edu.au Steve Cassidy Macquarie University steve.cassidy@mq.edu.au Abstract We give a hands-on demonstration of the Alveo Virtual Laboratory, a new plat- form for collaborative research in human communication science (HCS). Funded by the Australian Government National eResearch Collaboration Tools and Re- sources (NeCTAR) program, Alveo in- volves partners from a range of disci- plines: linguistics, natural language pro- cessing, speech science, psychology, as well as music and acoustic processing. The goal of the platform is to provide easy access to a variety of databases and a range of analysis tools, in order to fos- ter inter-disciplinary research and facili- tate the discovery of new methods for solving old problems or the application of known methods to new datasets. Alveo integrates a number of tools and enables non-technical users to process communi- cation resources (including not only text and speech corpora but also music re- cordings and videos) using these tools in a straightforward manner. 1 Introduction Alveo provides easy access to a range of data- bases relevant to human communication science disciplines, including speech, text, audio and video, some of which would previously have been difficult for researchers to access or even know about. The system implements a uniform and secure license management system for the diverse licensing and user agreement conditions required. Browsing, searching and dataset ma- nipulation are also functionalities which are available in a consistent manner across the data collections through the web-based Discovery Interface. 2 Alveo Tools and Corpora The first phase of the project, from December 2012 to June 2014 (Estival et al. 2013) saw the inclusion of data collections contributed by the project partners (see the list of partners in the Acknowledgments section). Some of these were already well-known, e.g. 1, 3 and 9, but some had been difficult of access or not available, e.g. 2, 5, 6, 7, 8. 1. PARADISEC (Pacific and Regional Archive for Digital Sources in Endangered Cultures: http://paradisec.org.au): audio, video, text and image resources for Australian and Pacific Island languages (Thieberger et al. 2011). 2. AusTalk (https://austalk.edu.au): audio-visual speech corpus of Australian English (Burnham et al. 2011). 3. The Australian National Corpus (https://www.ausnc.org.au) (Cassidy et al. 2012) comprising: Australian Corpus of English (ACE); Australian Radio Talkback (ART); Aus- tLit; Braided Channels; Corpus of Oz Early Eng- lish (COOEE); Griffith Corpus of Spoken Eng- lish (GCSAusE); International Corpus of English (ICE-AUS); Mitchell & Delbridge corpus; Monash Corpus of Spoken English (Musgrave and Haugh 2009). 4. AVOZES, a visual speech corpus (Goecke and Millar 2004). 5. UNSW Pixar Emotional Music Excerpts: Pix- ar movie theme music expressing different emo- tions. 6. Sydney University Room Impulse Responses: environmental audio samples which, through convolution with speech or music, can create the effect of that speech or music in that acoustic environment. 7. Macquarie University Battery of Emotional Prosody: sung sentences with different prosodic patterns. 8. Colloquial Jakartan Indonesian corpus: audio and text, recorded in Jakarta in the early 1990’s (ANU). 9. …

Content truncated to 500 words