| from datasets import load_dataset, concatenate_datasets, DatasetDict | |
| import pandas as pd | |
| current_dataset = load_dataset("muzaffercky/kurdish-kurmanji-voice-corpus") | |
| new_dataset = load_dataset("audiofolder", data_dir="./prepare_data") | |
| metadata_df = pd.read_csv("metadata.csv", sep=";") | |
| def add_transcription(example): | |
| path = example["audio"]["path"] | |
| *args, root_folder, folder, filename = path.split("/") | |
| full_name = f"{root_folder}/{folder}/{filename}" | |
| matching_rows = metadata_df[metadata_df["file_name"] == full_name] | |
| transcription = matching_rows["transcription"].values[0] | |
| source = matching_rows["source"].values[0] | |
| example["transcription"] = transcription | |
| example["url"] = source | |
| print(transcription) | |
| return example | |
| def merge_datasets(current: DatasetDict, new: DatasetDict): | |
| keys = set(current.keys()).union( | |
| set(new.keys()) | |
| ) | |
| updated_splits = {} | |
| for key in keys: | |
| current_has_key = key in current | |
| new_has_key = key in new | |
| if current_has_key and new_has_key: | |
| current_split = current[key] | |
| new_split = new[key] | |
| updated_splits[key] = concatenate_datasets( | |
| [current_split, new_split] | |
| ) | |
| elif current_has_key: | |
| updated_splits[key] = current[key] | |
| elif new_has_key: | |
| updated_splits[key] = new[key] | |
| return DatasetDict(updated_splits) | |
| new_dataset = new_dataset.map(add_transcription).remove_columns("label") | |
| updated_dataset = merge_datasets(current_dataset, new_dataset) | |
| updated_dataset.push_to_hub("muzaffercky/kurdish-kurmanji-voice-corpus") | |