Alexa Wake Words in EU Spanish (Adults)

Home » Case Study » Alexa Wake Words in EU Spanish (Adults)

Project Overview:

Objective

We successfully built a comprehensive dataset of audio clips featuring the “Alexa” wake word, as articulated in Mexican Spanish by adults. This dataset is now a pivotal asset for enhancing wake word recognition systems and voice assistants, specifically designed for this demographic group.

Scope

Our team gathered an extensive array of audio recordings from adult Mexican Spanish speakers. These recordings, captured in various settings and accents, were meticulously annotated by us to include precise wake word annotations, showcasing our expertise in data annotation.

Sources

Voice Assistant Users: Collaborate with Alexa users who are willing to contribute audio clips of them saying “Alexa” in different contexts.
Voice Actors: Hire professional voice actors to create synthetic wake word recordings in Mexican Spanish for added diversity.
Public Domain Recordings: Extract publicly available audio recordings with instances of the “Alexa” wake word in Mexican Spanish.

Data Collection Metrics

Total Audio Clips Collected and Annotated: 50,000 clips
User Contributions: 30,000
Voice Actor Recordings: 12,000
Public Domain Extracts: 8,000

Annotation Process

Stages

Wake Word Annotation: We accurately marked the start and end points of the “Alexa” wake word in each audio clip.
Speaker Demographics: We compiled metadata about each contributor, including age, accent, and gender.
Recording Conditions: We documented the conditions of each recording, such as background noise levels and the types of recording devices used.

Annotation Metrics

Audio Clips with Wake Word Annotations: 40,000
Speaker Demographics: 40,000
Recording Condition Metadata: 40,000

Quality Assurance

Stages

Annotation Verification: We conducted a thorough validation of the annotations using both automated tools and human reviewers.
User Consent and Privacy Compliance: We ensured all user-contributed audio clips were used with explicit consent and any personal information was anonymized. Our process strictly adhered to privacy regulations, allowing data contributors to opt-out or request data removal.

QA Metrics

Annotation Validation Cases: 4,000 (10% of total)
Privacy Audits: 25,000 (for user-contributed data)

Conclusion

As a leading data collection and annotation company, we are proud to present the Alexa Wake Words Dataset in Mexican Spanish (Adults). This dataset exemplifies our commitment to delivering high-quality, diverse, and accurately annotated datasets, essential for advancing voice recognition and natural language processing technologies.