README.md

metadata

pretty_name: ECD-TSE
size_categories:
  - 100M<n<1B
license: cc-by-nc-4.0
task_categories:
  - text-to-speech
language:
  - en
tags:
  - emotion
gated: true
extra_gated_prompt: >-
  Subject to other terms:

  1. The corpus is not used for commercial purposes and is only provided free of
  charge to “universities and research institutes” for scientific research.

  2. When publishing papers and applying for results, if you use this dataset,
  please indicate the reference:

     @article{liu2025towards,
       title={Towards Emotionally Consistent Text-Based Speech Editing: Introducing EmoCorrector and The ECD-TSE Dataset},
       author={Liu, Rui and Gao, Pu and Xi, Jiatian and Sisman, Berrak and Busso, Carlos and Li, Haizhou},
       journal={arXiv preprint arXiv:2505.20341},
       year={2025}
     }

  3. The final interpretation of this corpus belongs to S2LAB Lab, Inner
  Mongolia University, China.
extra_gated_fields:
  First Name: text
  Last Name: text
  Date of birth: date_picker
  Country: country
  Institution:
    type: text
    placeholder: e.g., Stanford University
    description: >-
      Please enter the full name of your institution (e.g., including
      'University' or 'Institute').
  Job title:
    type: select
    options:
      - Student
      - Research Graduate
      - AI researcher
      - AI developer/engineer
      - Reporter
      - Other
  geo: ip_location
  acceptance:
    type: checkbox
    label: >-
      By clicking submit below, I accept the terms of the license and
      acknowledge that the information I provide will be collected, stored and
      processed by S2LAB
extra_gated_description: The information you provide will be collected, stored and processed by S2LAB.
extra_gated_button_content: Submit
extra_gated_eu_disallowed: true

ECD-TSE

ECD-TSE Overview

The prominent aspect of ECD-TSE is its inclusion of paired data featuring diverse text variations and a range of emotional expressions. ECD-TSE encompasses several topics.

Attribute	Value
Speech Samples	84,000 = 1,400 * 5 * 12
Emotions	5 (Happy, Sadness, Neutral, Angry, Fear)
Speakers	12 (6 male and 6 female)
Text	7,000 = 1,400 * 5
Total Duration	Approximately 90 hours
Sampling Rate	16,000 Hz

Speakers	TTS Model	Original Dataset
Speaker 1-5	Azure
Speaker 6-10	Cosyvoice	ESD
Speaker 11-12	F5-TTS	MEAD

Topics	Description
Daily Life	Sentences describing everyday activities, habits, and common events
Social Interaction	Sentences involving communication, dialogue, and interaction between people.
Natural Environment	Sentences describing natural phenomena, plants, animals, and environmental features.
Narrative/Storytelling	Sentences likely extracted from novels, fairy tales, or other narrative texts, containing plots and characters.
Telephone Communication	Including customer service, business calls, etc.
News Broadcast	Current affairs news, weather forecasts, etc.
Meeting Minutes	Discussions, summaries, decisions, etc.
Education	Teaching materials, learning materials, language learning dialogues, etc.
Technology	Including technical terminology, product introductions, research discussions, industry trends, etc.
Medicine	Covering medical terminology, diagnoses, treatment plans, patient communication, etc.
Law	Involving legal provisions, case analyses, legal consultations, court proceedings, etc.

Citations

If you find this dataset useful in your research, please consider citing the following paper:

@article{liu2025towards,
  title={Towards Emotionally Consistent Text-Based Speech Editing: Introducing EmoCorrector and The ECD-TSE Dataset},
  author={Liu, Rui and Gao, Pu and Xi, Jiatian and Sisman, Berrak and Busso, Carlos and Li, Haizhou},
  journal={arXiv preprint arXiv:2505.20341},
  year={2025}
}