Map scheme labels

The labels attribute of schemes can be used to encode additional information about the table data. In the following example we have a scheme "transcription" that maps IDs to words. And a scheme "speaker" that holds gender and age information about the speakers in the database.

import audformat.testing

db = audformat.testing.create_db(minimal=True)
db.schemes["transcription"] = audformat.Scheme(
    labels={
        0: "hello",
        1: "goodbye",
    }
)
db.schemes["speaker"] = audformat.Scheme(
    labels={
        "spk1": {
            "gender": "male",
            "age": 33,
        },
        "spk2": {
            "gender": "female",
            "age": 30,
        },
        "spk3": {
            "gender": "male",
            "age": 37,
        },
    }
)
audformat.testing.add_table(
    db,
    "files",
    audformat.define.IndexType.FILEWISE,
)

If we request the transcription column, we get a pandas.Series with the word IDs:

db["files"]["transcription"].get()

	transcription
file
audio/001.wav	1
audio/002.wav	0
audio/003.wav	0
audio/004.wav	0
audio/005.wav	1

But if we are interested in the actual transcribed words, we can use the map argument to request them.

db["files"]["transcription"].get(map="transcription")

	transcription
file
audio/001.wav	goodbye
audio/002.wav	hello
audio/003.wav	hello
audio/004.wav	hello
audio/005.wav	goodbye

Note that we can pass any string to map. It will be used as the name of the returned pandas.Series.

db["files"]["transcription"].get(map="word")

	word
file
audio/001.wav	goodbye
audio/002.wav	hello
audio/003.wav	hello
audio/004.wav	hello
audio/005.wav	goodbye

Likewise, if we request the speaker column, a list of names is returned:

db["files"]["speaker"].get()

	speaker
file
audio/001.wav	spk2
audio/002.wav	spk1
audio/003.wav	spk1
audio/004.wav	spk3
audio/005.wav	spk3

If we are interested in the age of the speakers, we can do:

db["files"]["speaker"].get(map="age")

	age
file
audio/001.wav	30
audio/002.wav	33
audio/003.wav	33
audio/004.wav	37
audio/005.wav	37

This also works for tables. Here we pass a dictionary with column names as keys and scheme fields as values.

map = {
    "speaker": "age",
}
db["files"].get(map=map)

	transcription	age
file
audio/001.wav	1	30
audio/002.wav	0	33
audio/003.wav	0	33
audio/004.wav	0	37
audio/005.wav	1	37

It is possible to map several columns at once and to map the same column to multiple fields.

map = {
    "transcription": "words",
    "speaker": ["age", "gender"],
}
db["files"].get(map=map)

	words	age	gender
file
audio/001.wav	goodbye	30	female
audio/002.wav	hello	33	male
audio/003.wav	hello	33	male
audio/004.wav	hello	37	male
audio/005.wav	goodbye	37	male

To keep the original columns values, we can include the column name in the list.

map = {
    "transcription": ["transcription", "words"],
    "speaker": ["speaker", "age", "gender"],
}
db["files"].get(map=map)

	speaker	transcription	words	age	gender
file
audio/001.wav	spk2	1	goodbye	30	female
audio/002.wav	spk1	0	hello	33	male
audio/003.wav	spk1	0	hello	33	male
audio/004.wav	spk3	0	hello	37	male
audio/005.wav	spk3	1	goodbye	37	male