Adding Speaker Demographics

We’ve tried to make adding speaker demographics to fave-extract output as flexible as possible, including

File Formats

Excel or CSV files

To ensure demographic information in a an .xlsx or .csv file is correctly included in fave-extract output two columns are required:

Required Columns

  • file_name: The file stem of the wav and textgrid files

  • speaker_num: The speaker to be analyzed in a file. the first speaker is 1.

So, if you had a corpus that looked like this:

../my_corpus
├── speaker1.TextGrid
├── speaker1.wav
├── speaker2.TextGrid
└── speaker2.wav

Your excel file or csv file would have to look something like this:

file_name speaker_num age
speaker1 1 26
speaker2 1 50
speaker2 2 23
Tip

If a speaker demographics file is provided, fave-extract will only process data for speakers with entries.

YAML file

Another option for formatting speaker demographic information is in a yaml file. Yaml is a very flexible data structuring format. For this corpus:

../my_corpus
├── speaker1.TextGrid
├── speaker1.wav
├── speaker2.TextGrid
└── speaker2.wav

An speaker demographics yaml file would look like

# yaml
- file_name: speaker1
  speaker_num: 1
  age: 26
- file_name: speaker3
  speaker_num: 1
  age: 50
- file_name: speaker1
  speaker_num: 1
  age: 23
Required Fields

The file_name and speaker_num fields are required.

Flexibility

Outside of the required fields

  • Not every speaker has to have the same fields defined.

  • The fields don’t need to appear in a consistent order.

Legacy-fave speaker file

If you have legacy-fave .speaker files, you can pass them to the --speakers option.

Usage

All three fave-extract subcommands support passing of demographic files.

fave-extract corpus my_corpus/ --speakers demographics.csv