CHINESE QUEER CORPUS

CQC · 中文酷儿语料库 · Mandarin speech, sex, gender, and sexuality

Data Use Agreement

Chinese Queer Corpus (CQC) Data Use Agreement, version 1.0

This page sets out the terms for prospective applicants. After approval, the executed agreement is sent for signature by the Custodian and an authorized representative of the Recipient’s Institution.

This Agreement is between the CQC Research Team as data custodian (“Custodian”) and the Recipient named in the signed application, together with the Recipient’s institution (“Institution”).

1. Definitions

  1. Speaker: any person whose voice appears in a CQC source recording.
  2. Public Materials: this website, the datasheet, annotation scheme, and aggregate statistics. These are not covered by this Agreement.
  3. Controlled Data: transcripts, TextGrids, speaker- and session-level metadata (including identity fields), timestamps, and audio files.
  4. Restricted Derived Data: any data created from Controlled Data at the level of individual speakers, sessions, segments, or frames (for example per-speaker acoustic measurements or re-annotations), and any data that could identify a Speaker or reconstruct Controlled Data.
  5. Aggregate Results: group-level statistics, model estimates, tables, and figures of distributions that meet Section 5.1. Aggregate Results are not Restricted Derived Data.
  6. Authorized Users: the Recipient and the named team members listed in the application, each of whom has read this Agreement.

2. Permitted use

  1. Controlled Data may be used only for the non-commercial research project described in the application.
  2. Permitted purposes include phonetic, sociophonetic, linguistic, and social-scientific analysis, and the evaluation of speech technology performance across speaker groups in aggregate.
  3. A new project or a substantially changed purpose requires a new application.

3. Prohibited use

Authorized Users shall not:

  1. Infer, predict, or classify any Speaker’s gender identity, trans status, sex characteristics, sexual orientation, or other identity attribute from speech, text, or metadata, or train or evaluate any system for that purpose.
  2. Identify or attempt to identify any Speaker, including by linking Controlled Data with other datasets or applying speaker recognition against external data.
  3. Contact any Speaker in connection with CQC.
  4. Use any Speaker’s voice for voice cloning, speech synthesis, voice conversion, or the training of generative speech models.
  5. Use Controlled Data for commercial purposes, surveillance, law enforcement, immigration control, content moderation targeting individuals, or any decision about an individual.
  6. Redistribute, sublicense, sell, or publish Controlled Data or Restricted Derived Data, including by uploading them to public repositories or model hubs, or by entering them into third-party AI services that retain data.
  7. Publish anything that discloses, or could reasonably reveal, a Speaker’s identity together with an identity attribute.

4. Security

  1. Controlled Data and Restricted Derived Data shall be stored on encrypted, access-controlled systems at the Institution and accessed only by Authorized Users.
  2. Recipient shall notify the Custodian within 72 hours of any suspected unauthorized access, use, or disclosure, and cooperate in remediation.

5. Publication

  1. Aggregate Results may be published. Group-level statistics shall not be reported for groups of fewer than 3 speakers unless the Speakers cannot be distinguished.
  2. Audio clips, and spectrograms, waveforms, or pitch tracks of an individual Speaker’s speech, require prior written permission from the Custodian, because such figures can be used to recover or recognize a voice. Figures of aggregate distributions need no permission.
  3. Verbatim quotations longer than 20 Chinese characters require prior written permission, because longer quotations can be searched to locate source videos.
  4. Speakers shall be referred to only by CQC speaker IDs.
  5. Publications shall cite the CQC release used, as described on the About page, and state that the data were used under this Agreement.

6. Updates and removal

  1. The Custodian removes material when a Speaker requests it or when a source becomes unavailable, and issues updated releases.
  2. Within 30 days of a removal notice, Recipient shall delete the affected Controlled Data and any Restricted Derived Data containing it, and confirm deletion in writing. Aggregate Results already published need not be withdrawn.
  3. Recipient shall use the most recent release for new analyses.

7. Term and termination

  1. This Agreement runs for 3 years from approval and may be renewed.
  2. At the end of the term or the project, or upon breach, Recipient shall delete all Controlled Data and Restricted Derived Data and confirm in writing. Aggregate Results may be kept.
  3. The Custodian may suspend access and notify the Recipient’s Institution of a breach.

8. General

  1. Controlled Data are provided as is, without warranty of accuracy or fitness for any purpose.
  2. Recipient is responsible for obtaining any ethics approval required by its Institution and for complying with the laws that apply to it.
  3. This Agreement does not transfer any copyright in source recordings, which remains with the rights holders.
  4. The executed agreement specifies governing law and the authorized signatories, and prevails over this page.