MELD_audio.py

#!/usr/bin/env python3
# -*- coding: utf-8 -*-

import datasets
import pandas as pd
from datasets import ClassLabel
import os

"""The Audio, Speech, and Vision Processing Lab - Emotional Sound Database (ASVP - ESD)"""

_HOMEPAGE = "https://affective-meld.github.io/"

_CITATION = """\
@article{poria2018meld,
  title={Meld: A multimodal multi-party dataset for emotion recognition in conversations},
  author={Poria, Soujanya and Hazarika, Devamanyu and Majumder, Navonil and Naik, Gautam and Cambria, Erik and Mihalcea, Rada},
  journal={arXiv preprint arXiv:1810.02508},
  year={2018}
}
@article{chen2018emotionlines,
  title={Emotionlines: An emotion corpus of multi-party conversations},
  author={Chen, Sheng-Yeh and Hsu, Chao-Chun and Kuo, Chuan-Chun and Ku, Lun-Wei and others},
  journal={arXiv preprint arXiv:1802.08379},
  year={2018}
}
"""

_DESCRIPTION = """\
Multimodal EmotionLines Dataset (MELD) has been created by enhancing and extending EmotionLines dataset. 
MELD contains the same dialogue instances available in EmotionLines, but it also encompasses audio and 
visual modality along with text. MELD has more than 1400 dialogues and 13000 utterances from Friends TV series. 
Multiple speakers participated in the dialogues. Each utterance in a dialogue has been labeled by any of these 
seven emotions -- Anger, Disgust, Sadness, Joy, Neutral, Surprise and Fear. MELD also has sentiment (positive, 
negative and neutral) annotation for each utterance.
This dataset is modified from https://huggingface.co/datasets/zrr1999/MELD_Text_Audio.
The audio is extracted from MELD mp4 files while the audio only has one channel with sample rate 16khz.
"""

_LICENSE = "gpl-3.0"


class MELD_Audio(datasets.GeneratorBasedBuilder):

    DEFAULT_WRITER_BATCH_SIZE = 256
    DEFAULT_CONFIG_NAME = "MELD_Audio"
    BUILDER_CONFIGS = [  # noqa: RUF012
        datasets.BuilderConfig(name="MELD_Audio", version=datasets.Version("0.0.1"), description="MELD audio"),
    ]

    def _info(self):
        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=datasets.Features(
                {
                    "text": datasets.Value("string"),
                    "path": datasets.Value("string"),
                    "audio": datasets.Audio(sampling_rate=16000),
                    "emotion": datasets.Value("string"),  # ClassLabel(names=["neutral", "joy", "sadness", "anger", "fear", "disgust", "surprise"]),
                    "sentiment": datasets.Value("string"), # ClassLabel(names=["neutral", "positive", "negative"]),
                }
            ),
            homepage=_HOMEPAGE,
            license=_LICENSE,
            citation=_CITATION,
        )

    def _split_generators(self, dl_manager):
        """Returns SplitGenerators."""
        metadata_dir: dict[str, str] = dl_manager.download_and_extract(
            {"train": "train.csv", "dev": "dev.csv", "test": "test.csv"}
        )  # type: ignore  # noqa: PGH003
        data_path: dict[str, str] = dl_manager.download(
            {
                "train": "archive/train.tar.gz",
                "dev": "archive/dev.tar.gz",
                "test": "archive/test.tar.gz",
            }
        )  # type: ignore  # noqa: PGH003
        local_extracted_archive: dict[str, str] = (
            dl_manager.extract(data_path)
            if not dl_manager.is_streaming
            else {
                "train": None,
                "dev": None,
                "test": None,
            }
        )  # type: ignore  # noqa: PGH003

        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                gen_kwargs={
                    "filepath": metadata_dir["train"],
                    "split": "train",
                    "local_extracted_archive": local_extracted_archive["train"],
                    "audio_files": dl_manager.iter_archive(data_path["train"]),
                },
            ),
            datasets.SplitGenerator(
                name=datasets.Split.VALIDATION,
                gen_kwargs={
                    "filepath": metadata_dir["dev"],
                    "split": "dev",
                    "local_extracted_archive": local_extracted_archive["dev"],
                    "audio_files": dl_manager.iter_archive(data_path["dev"]),
                },
            ),
            datasets.SplitGenerator(
                name=datasets.Split.TEST,
                gen_kwargs={
                    "filepath": metadata_dir["test"],
                    "split": "test",
                    "local_extracted_archive": local_extracted_archive["test"],
                    "audio_files": dl_manager.iter_archive(data_path["test"]),
                },
            ),
        ]

    def _generate_examples(self, filepath, split, local_extracted_archive, audio_files):
        """Yields examples."""
        metadata_df = pd.read_csv(filepath, sep=",", index_col=0, header=0)
        metadata = {}
        for _, row in metadata_df.iterrows():
            id_ = f"dia{row['Dialogue_ID']}_utt{row['Utterance_ID']}"
            audio_path = f"{split}/{id_}.flac"
            metadata[audio_path] = row

        id_ = 0
        for path, f in audio_files:
            if path in metadata:
                row = metadata[path]
                path = os.path.join(local_extracted_archive, path) if local_extracted_archive else path
                audio = {"path": path, bytes: f.read()}
                yield (
                    id_,
                    {
                        "text": row["Utterance"],
                        "path": path,
                        "audio": audio,
                        "emotion": row["Emotion"],
                        "sentiment": row["Sentiment"],
                    },
                )
                id_ += 1