The first part of this year I am concentrating on coming to grips with a new set of tools for managing workflows. Specifically the Galaxy environment and Docker, a platform for deploying self contained virtual environments. The aim of the project is to distribute highly customizable and flexible workflows for interacting with speech and associated labels. This will be mainly be achieved using tools developed in R but the beauty of Galaxy is that you can string together tools from a variety of sources (e.g. Python or C), cobbling them together into a workflow that can be tailored to the needs of the user.
Galaxy was developed for the bioinformatics market and has been very influential in the progress of genetics research as highly computationally intensive work can be distributed and replicated quickly and efficiently. Tools and Wokflows can be shared and customized. In recent times Galaxy has been re-purposed as a way of working with all manners of medium to big data. This is the platform chosen by Alveo to interact with Austalk and related speech and text corpora. This can be thought of as a virtual laboratory which produces replicable, citable and publishable research based on primary data.
My current project is to find an efficient way to handle relatively long recordings and quickly segmenting and labelling these, either manually or automatically, at a word and phoneme level. These can then be phonetically analysed more closely and also provides a way of searching audio at a level other than the orthographic word. The long-term plan is to use Galaxy along with a set of R tools to parse files in a variety of labelling standards and most importantly keep them time-aligned with their associated audio files. I will see if I can write a semi-regular blog of my exploits and then I will aggregate that into something a little more readable – and hopefully publishable – in the Articles section of this site.