
Audio transcription
Context
Research Objectives
Insights informed
The language environment around a growing infant is rich. Parents love talking to their children and interact with them. However, analyzing the audio data of these interactions could prove to be a challenge due to the sheer volume of data (4000 hours in this case).
-
Identify and compile an audio annotation scheme that fits the bill
-
Create a document with instructions on how to complete an annotation task with minimal support
Efficient data analyses of a large corpus of data and guidelines for current and future research assistants.
Research Questions
What is the best method of transcribing audio data to fit the research needs of the lab?
How do we create a module that has a strong foundation upon which layers can be added?
How do we package the information so it is accessible to different users?
Methodologies
Literature reviews
Expert interviews
Process & Challenges
ELAN audio software
We set out by exploring what annotation schemes already exist in the field. After this, it was about narrowing down upon what made most sense for the research questions we were trying to answer in our lab.
Challenges:
1. Expert interviews: Experts were across the world and logistically there were several constraints in scheduling calls and coordinating meetings.
2. Vague problem space: Apart from the research topics of the lab, everything had to be defined by detailed interviews with stakeholders.
Impact
-
An annotation scheme that is replicable, easy to learn and modify
-
A document that can be used by current and future members of the research lab to analyze huge corpuses of data that are collected