AusKidTalk workflow

In the AusKidTalk corpus, speech data was collected from 620 children aged 3-12 using five tasks. To provide accurate orthographic transcription for the single word production task (Task1), we developed a workflow concatenating multiple automatic speech processing tools and augmented it with hand-correction. For some files, automatic speech processing was not possible, therefore, these files were transcribed manually, from scratch.

Three workflows are provided for

Evaluating the quality of the audio recording and the automatically generated orthographic transcription

Hand-correcting automatically generated orthographic transcription

Creating orthographic transcription from scratch, when

Input data

Input data are not provided to protect participants' privacy; however, data are available for research purposes upon request .

Workflows

Workflow 1: AusKidTalk_Task1_prescreening.zip

This workflow loads a wav file with a matchin textgrid and prompts annotators to evaluate data quality.

Workflow 2: AusKidTalk_Task1_hand_correction.zip

This workflow loads a wav file with a matchin textgrid and prompts annotators to correct automatically generated orthographic transcription.

Workflow 3: AusKidTalk_Task1_transcription_from_scratch.zip

This workflow assists annotators in manually identifying Task(1) in a wav file, then assits them in creating time-aligned orthographic transcription from scratch.