[BUG] Wrong columns names in crowd speech dataset about crowd-kit

[dustalov]

Hi, thank you for reporting this! I can confirm it is a bug.

The load_crowdspeech_dataframes() method does not renames the corresponding fields: https://github.com/Toloka/crowd-kit/blob/main/crowdkit/datasets/_loaders.py#L84

You can work around this bug by manually renaming the performer column to the worker column and text to output.

Also, please add the title to this reported issue.

[pilot7747]

@dustalov I maybe blind but I see
labels = pd.read_csv(join(data_path, 'crowd_labels.csv')).rename(columns={'output': 'text', 'performer': 'worker'})

Seems like fields names must be correct.

[pilot7747]

@Senarect I checked on my machine an on crowdkit==1.2.0 everything seems correct.

[pilot7747]

@Senarect Pls update the Crowd-Kit and inform us if the issue still appears.

[dustalov]

@pilot7747 @Senarect I think I understand the confusion.

In CrowdSpeech currently we have task, worker, and text columns. However, some implementation of our algorithms for textual aggregation require output column, some require text column, see #61.

If we should keep text, please close this issue.

[pilot7747]

Closing the issue in favour of fixing the bug in #61