Dataset Card for ciempiess_complementary

Dataset Summary

The CIEMPIESS COMPLEMENTARY is a phonetically balanced corpus of isolated Spanish words spoken by people of Central Mexico. It was designed to solve one particular issue when training automatic speech recognition (ASR) systems in the Spanish of Central Mexico. This problem appears when someone collects some training data, but the system complains because it does not find enough instances of one or more particular phoneme.

The CIEMPIESS COMPLEMENTARY Corpus was created with the voices of 10 male and 10 female volunteers reading isolated words. The words were chosen to assure users to get, at least, twenty instances of every single phoneme and allophone of the Mexican phonetic alphabet called Mexbet. So, the CIEMPIESS COMPLEMENTARY is "COMPLEMENTARY" because it "complements" datasets when training ASR systems in the Spanish of Central Mexico.

On the other hand, Mexbet is a phonetic alphabet designed for the Spanish of Central Mexico. It was created in 2004 by the linguist Javier Cuétara at UNAM University to do experiments with another corpus in Mexican Spanish called DIMEx100, giving excellent results. For these reasons, Mexbet is the preferred alphabet for the all CIEMPIESS family. An updated set of the of the Mexbet rules adapted for the CIEMPIESS project is presented in this paper.

The CIEMPIESS COMPLEMENTARY is part of the CIEMPIESS Experimentation, which is a set of three different datasets, specifically CIEMPIESS COMPLEMENTARY, CIEMPIESS FEM and CIEMPIESS TEST.

CIEMPIESS is the acronym for:

"Corpus de Investigación en Español de México del Posgrado de Ingeniería Eléctrica y Servicio Social".

Example Usage

The CIEMPIESS COMPLEMENTARY contains only the train split:

from datasets import load_dataset
ciempiess_complementary = load_dataset("ciempiess/ciempiess_complementary")

It is also valid to do:

from datasets import load_dataset
ciempiess_complementary = load_dataset("ciempiess/ciempiess_complementary",split="train")

Supported Tasks

automatic-speech-recognition: The dataset can be used to test a model for Automatic Speech Recognition (ASR). The model is presented with an audio file and asked to transcribe the audio file to written text. The most common evaluation metric is the word error rate (WER).

Languages

The language of the corpus is Spanish with the accent of Central Mexico.

Dataset Structure

Data Instances

{
  'audio_id': 'CMPC_F_05_A_0002', 
  'audio': {
    'path': '/home/carlos/.cache/HuggingFace/datasets/downloads/extracted/7c889e5ab17b842a7b62a3fbcfe0f9d71c0b89cb8df5a896a9acae10594bffba/train/female/F_05/alphabet/CMPC_F_05_A_0002.flac', 
    'array': array([0.        , 0.        , 0.        , ..., 0.00097656, 0.00183105,
       0.00180054], dtype=float32), 'sampling_rate': 16000
  }, 
  'speaker_id': 'F_05', 
  'gender': 'female', 
  'duration': 9.470000267028809, 
  'utt_type': 'Letters', 
  'age': 20, 
  'education': 'University', 
  'birthplace': 'Mexico City', 
  'residence': 'Mexico City', 
  'normalized_text': 'jota ka ele doble ele eme ene eñe o pe cu erre'
}

Data Fields

• audio_id (string) - id of audio segment
• audio (datasets.Audio) - a dictionary containing the path to the audio, the decoded audio array, and the sampling rate. In non-streaming mode (default), the path points to the locally extracted audio. In streaming mode, the path is the relative path of an audio inside its archive (as files are not downloaded and extracted locally).
• speaker_id (string) - id of speaker.
• gender (string) - gender of speaker (male or female).
• duration (float32) - duration of the audio file in seconds.
• utt_type (string) - utterance type. "W": words, "A": letters (Alphabet), "D": digits.
• age (int32) - age of the speaker.
• education (string) - educationale level of the speaker.
• birthplace (string) - birthplace of the speaker.
• residence (string) - place of residence of the speaker.
• normalized_text (string) - normalized audio segment transcription.

Data Splits

The corpus counts just with the train split which has a total of 520 speech files from 10 male and 10 female volunteers from Central Mexico with ages between 20 to 49 years old. Each speaker contributes with 26 speech files.

Dataset Creation

Curation Rationale

The CIEMPIESS COMPLEMENTARY Corpus has the following characteristics:

• The corpus was recorded using a Sony recorder model ICD-PX312D in a moderate noise environment similar to a medium size library. The recordings were originally recorded in MP3 format with a quality of 44.1 khz, 128 kbps, stereo.

• 10 male and 10 female volunteers from Central Mexico and ages from 20 to 49 years contributed with 26 speech files each.

• The corpus is 56 minutes long with 520 speech files converted to a 16 kHz , 16-bit, PCM, mono format.

• Each of the speakers read the digits from zero to nine (1 speech file), the alphabet with some common nick names of certain letters, like "i griega" for the "y" (3 speech files, 11 letters per file) and finally, every speaker reads a list of 66 words (22 speech files, 3 words per file). In general, every speaker reads different words, but some few words are read by two different speakers.

• The 22 lists of Spanish words that every speaker reads are designed to ensure that the 66 phonemes and allophones of the T66 level of Mexbet are repeated, at least, one time for each speaker. Note that the 29 phonemes of the T29 level of Mexbet are included in the T66 level.

• Speakers in the CIEMPIESS COMPLEMENTARY are not present in any other CIEMPIESS dataset.

Source Data

Initial Data Collection and Normalization

All the recordings that constitute the CIEMPIESS COMPLEMENTARY were recorded with a Sony recorder model ICD-PX312D in a moderate noise environment similar to a medium size library by Susana Alejandra Jiménez Sandoval between 2016 and 2017. The corpus was curated by Carlos Daniel Hernández Mena in 2018.

Annotations

Annotation process

The corpus was created out of recordings of volunteers reading isolated words and letters. The lists of words are unique for each speaker and they were created to be phonetically balanced and to be read by native speakers of the Spanish language with the accent of Central Mexico.

Who are the annotators?

The corpus was designed and curated by Carlos Daniel Hernández Mena and it was collected by Susana Alejandra Jiménez Sandoval.

Personal and Sensitive Information

The corpus is comprised of speech recordings of isolated words and letters read by volunteers, so, no personal or sensitive information is shared. However, you agree to not attempt to determine the identity of speakers in this dataset.

Considerations for Using the Data

Social Impact of Dataset

This dataset is valuable because it is phonetically balanced.

Discussion of Biases

The corpus is gender and phonetically balanced; 10 male and 10 female volunteers from Central Mexico and ages from 20 to 49 years contributed with 26 speech files each.

Other Known Limitations

"CIEMPIESS COMPLEMENTARY CORPUS" by Carlos Daniel Hernández Mena is licensed under a Creative Commons Attribution-ShareAlike 4.0 International (CC BY-SA 4.0) License with the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

Dataset Curators

The dataset was curated by Carlos Daniel Hernández Mena in 2018.

Licensing Information

CC-BY-SA-4.0

Citation Information

@misc{carlosmenaciempiesscomplementary2019,
      title={CIEMPIESS COMPLEMENTARY CORPUS: Audio and Transcripts of Spanish Isolated Words.}, 
      ldc_catalog_no={LDC2019S07},
      DOI={https://doi.org/10.35111/xdx5-n815},
      author={Hernandez Mena, Carlos Daniel and Jiménez Sandoval, Susana Alejandra},
      journal={Linguistic Data Consortium, Philadelphia},
      year={2019},
      url={https://catalog.ldc.upenn.edu/LDC2019S07},
}

Contributions

The authors want to thank to Alejandro V. Mena, Elena Vera and Angélica Gutiérrez for their support to the social service program: "Desarrollo de Tecnologías del Habla."