soerenray/speech_commands_enriched_and_annotated

Dataset Summary

📊 Data-centric AI principles have become increasingly important for real-world use cases.
At Renumics we believe that classical benchmark datasets and competitions should be extended to reflect this development.

🔍 This is why we are publishing benchmark datasets with application-specific enrichments (e.g. embeddings, baseline results, uncertainties, label error scores). We hope this helps the ML community in the following ways:

1. Enable new researchers to quickly develop a profound understanding of the dataset.
2. Popularize data-centric AI principles and tooling in the ML community.
3. Encourage the sharing of meaningful qualitative insights in addition to traditional quantitative metrics.

📚 This dataset is an enriched version of the speech_commands dataset. It provides predicted labels, their annotations and embeddings, trained with Huggingface's AutoModel and AutoFeatureExtractor. If you would like to have a closer look at the dataset and model's performance, you can use Spotlight by Renumics to find complex sub-relationships between classes.

Explore the Dataset

The enrichments allow you to quickly gain insights into the dataset. The open source data curation tool Renumics Spotlight enables that with just a few lines of code:

Install datasets and Spotlight via pip:

!pip install renumics-spotlight datasets

Load the dataset from huggingface in your notebook:

import datasets

dataset = datasets.load_dataset("soerenray/speech_commands_enriched_and_annotated", split="train")

Start exploring with a simple view that leverages embeddings to identify relevant data segments:

from renumics import spotlight

df = dataset.to_pandas()
df_show = df.drop(columns=['embedding', 'logits'])
spotlight.show(df_show, port=8000, dtype={"audio": spotlight.Audio, "embedding_reduced": spotlight.Embedding})

You can use the UI to interactively configure the view on the data. Depending on the concrete tasks (e.g. model comparison, debugging, outlier detection) you might want to leverage different enrichments and metadata.

Speech commands Dataset

The speech commands dataset consists of 60,973 samples in 30 classes (with an additional silence class). The classes are completely mutually exclusive. It was designed to evaluate keyword spotting models. We have enriched the dataset by adding audio embeddings generated with a MIT's AST. Here is the list of classes in the speech commands:

Class
"Yes"
"No"
"Up"
"Down"
"Left"
"Right"
"On"
"Off"
"Stop"
"Go"
"Zero"
"One"
"Two"
"Three"
"Four"
"Five"
"Six"
"Seven"
"Eight"
"Nine"
"Bed"
"Bird"
"Cat"
"Dog"
"Happy"
"House"
"Marvin"
"Sheila"
"Tree"
"Wow"

Supported Tasks and Leaderboards

• TensorFlow Speech Recognition Challenge: The goal of this task is to build a speech detector. The leaderboard is available here.

Languages

English class labels.

Dataset Structure

Data Instances

A sample from the dataset is provided below:

{
  "audio": {
    "path":'bed/4a294341_nohash_0.wav', 
    "array": array([0.00126146 0.00647549 0.01160542 ... 0.00740056 0.00798924 0.00504583]), 
    "sampling_rate": 16000
    },
  "label": 20,  # "bed"
  "is_unknown": True,
  "speaker_id": "4a294341",
  "utterance_id": 0,
  "logits": array([-9.341216087341309, -10.528160095214844, -8.605941772460938, ..., -9.13764476776123, 
          -9.4379243850708, -9.254714012145996]),
  "Probability": 0.99669,
  "Predicted Label": "bed",
  "Annotated Labels": "bed",
  "embedding": array([ 1.5327608585357666, -3.3523001670837402, 2.5896875858306885, ..., 0.1423477828502655, 
  2.0368740558624268, 0.6912304759025574]),
  "embedding_reduced": array([-5.691406726837158, -0.15976890921592712])

}

Data Fields

Feature	Data Type
audio	Audio(sampling_rate=16000, mono=True, decode=True, id=None)
label	Value(dtype='int64', id=None)
is_unknown	Value(dtype='bool', id=None)
speaker_id	Value(dtype='string', id=None)
utterance_id	Value(dtype='int8', id=None)
logits	Sequence(feature=Value(dtype='float32', id=None), length=35, id=None)
Probability	Value(dtype='float64', id=None)
Predicted Label	Value(dtype='string', id=None)
Annotated Labels	Value(dtype='string', id=None)
embedding	Sequence(feature=Value(dtype='float32', id=None), length=768, id=None)
embedding_reduced	Sequence(feature=Value(dtype='float32', id=None), length=2, id=None)

Source Data

Initial Data Collection and Normalization

[More Information Needed]

Who are the source language producers?

[More Information Needed]

Annotations

Annotation process

[More Information Needed]

Who are the annotators?

[More Information Needed]

Personal and Sensitive Information

[More Information Needed]

Considerations for Using the Data

Social Impact of Dataset

[More Information Needed]

Discussion of Biases

[More Information Needed]

Other Known Limitations

[More Information Needed]

Additional Information

Dataset Curators

[More Information Needed]

Licensing Information

[More Information Needed]

Citation Information

If you use this dataset, please cite the following paper:

@article{speechcommandsv2,
   author = { {Warden}, P.},
    title = "{Speech Commands: A Dataset for Limited-Vocabulary Speech Recognition}",
  journal = {ArXiv e-prints},
  archivePrefix = "arXiv",
  eprint = {1804.03209},
  primaryClass = "cs.CL",
  keywords = {Computer Science - Computation and Language, Computer Science - Human-Computer Interaction},
    year = 2018,
    month = apr,
    url = {https://arxiv.org/abs/1804.03209},
}

Contributions

Pete Warden and Soeren Raymond(Renumics GmbH).