Alexa Wake Words in Mexican Spanish (Youth)

Home » Case Study » Alexa Wake Words in Mexican Spanish (Youth)

Project Overview:

Objective

To build a dataset of audio clips containing the “Alexa” wake word as spoken in Mexican Spanish by youth. This dataset will be essential for enhancing wake word detection systems and voice assistants designed for Mexican Spanish-speaking youth.

Scope

Our team embarked on an ambitious journey to amass a rich and varied collection of audio recordings. These recordings, contributed by youthful voices from diverse Mexican Spanish backgrounds, were meticulously curated to cover an array of accents and situational contexts. This collection not only exemplifies the diverse linguistic landscape of Mexican Spanish-speaking youth but also enhances the relatability and efficiency of voice recognition systems.

Sources

Participants: Collaborate with Mexican Spanish-speaking youth who consent to contribute audio clips of them saying “Alexa” in different contexts.
Voice Actors: Hire professional voice actors fluent in Mexican Spanish to create synthetic wake word recordings for added diversity and control.

Data Collection Metrics

Total Audio Clips: 15,000 clips
Participant Contributions: 9,000
Voice Actor Recordings: 6,000

Annotation Process

Stages

Wake Word Annotation: Accurately mark the temporal boundaries of the “Alexa” wake word within each audio clip.
Participant Demographics: Gather metadata about participants, including age, accent, and gender.
Recording Conditions: Document recording conditions such as ambient noise levels and recording devices used.

Annotation Metrics

Audio Clips with Wake Word Annotations: 15,000
Participant Demographic Metadata: 15,000
Recording Condition Metadata: 15,000

Quality Assurance

Stages

Annotation Verification: Implement a robust validation process involving automated verification tools and human reviewers to ensure precise wake word annotations.
User Consent: Ensure that participants’ audio clips have explicit consent for usage in the dataset and anonymize any personally identifiable information.
Privacy Compliance: Adhere to privacy regulations, including data protection policies and mechanisms for participants to opt out or request data removal.

QA Metrics

Annotation Validation Cases: 1,500 (10% of total)
Privacy Audits: 9,000 (for participant-contributed data)

Conclusion

The Alexa Wake Words Dataset in Mexican Spanish (Youth) aims to improve wake word detection technology and voice assistant systems tailored to Mexican Spanish-speaking youth. With diverse recordings, meticulous annotations, and privacy compliance, it serves as a valuable resource for voice recognition and natural language processing research and development.