ASR

This page provides an overview of how to use automatic speech recognition technology for linguistic transcription work. It also serves as the reference page for my software demo at ICAME42 - you can find all resources below.

Software Demo Steps

Prerequisites: You need an IBM Cloud account and have Python installed on your machine. You also need to install the library textgrids by Tommi Nieminen. More info can also be found in the script.

01. Get testfile: Download the testfile from above and store it somewhere on your machine where you can easily find it.

02. Get API&URL: Set up the Watson Speech-to-text service and copy your individual API & URL.

03. Get script: Copy/download the script Watson_STT_To_Textgrid and insert your API & URL. Add the directory of the testfile and specify an output directory.

01. Get testfile: Download the testfile from above and store it somewhere on your machine where you can easily find it.

02. BAS Web Services: Go to BAS Web Services and select “ASR“. Log-in with your institutional account or BAS user license and upload Testfile.wav.

03. Set up service options: Language: "English (Great Britain) (eml/google/lst/watson/webasr)" | ASR service: "IBM Watson ASR" | Output format: "Praat (TextGrid)" | Keep other default options.

05. Optional: Transform word level transcription into utterance level transcription with word_to_utterance_2.0.py script.

LaBB-CAT is a freely available browser-based corpus management and data mining tool developed by Robert Fromont and Jen Hay at the New Zealand Institute of Language, Brain and Behaviour. You can find the corresponding article here. LaBB-CAT is well documented and provides a great overview page as well as detailed step-by-step video tutorials.