Mayo Speech-to-Text: Bringing an Endangered Language into the Digital Age

Mayo, or Yoreme nooki, is a Uto-Aztecan language spoken primarily in the Mexican states of Sonora and Sinaloa. With around 40,000 speakers, it is an endangered language that has survived centuries of pressure from Spanish. Today, technology offers new opportunities for preservation and revitalization. Accurate speech-to-text for Mayo can help communities document oral histories, create educational materials, and produce subtitled media. However, building a reliable ASR system for Mayo comes with unique challenges.

Why Mayo Speech-to-Text Matters

For indigenous communities, language is a cornerstone of identity. Mayo speakers are often bilingual in Spanish, but the language is still used in homes, ceremonies, and traditional storytelling. Transcribing Mayo audio allows these voices to be captured in writing, making them accessible to younger generations who may be more comfortable reading. It also supports linguistic research, helping scholars study the structure of this underdocumented language. Moreover, subtitles in Mayo can make video content more inclusive for speakers with hearing impairments or low Spanish literacy.

Specific Transcription Challenges

Mayo presents several obstacles for automatic speech recognition:

• Vowel length: Contrasts like /a/ vs /aː/ can change word meaning. For example, 'chuki' (rain) versus 'chuuki' (squirrel). Generic ASR often conflates these.
• Glottal stop: The phoneme /ʔ/ is common in Mayo but absent in Spanish. It is often omitted or misheard by models trained on Spanish or English.
• Code-switching: Most Mayo conversations mix Spanish words and phrases. A good ASR must handle both languages seamlessly.
• Lack of training data: There are very few publicly available Mayo audio-transcript pairs. Traditional supervised learning is difficult.

Speechyou tackles these issues with a combination of transfer learning from related Uto-Aztecan languages (like Yaqui) and an active learning pipeline that improves with user-provided corrections. The model is fine-tuned to recognize vowel duration and glottal stops, and it can output text in a standard Latin orthography that marks length with double vowels.

Use Cases in Practice

• Oral history preservation: Record elders telling stories in Mayo, then transcribe them automatically. The text can be archived and searched for keywords.
• Bilingual education: Teachers can upload recordings of Mayo lessons and generate subtitles for videos shown in class. This helps students connect spoken and written forms.
• Community media: Local radio stations can add Mayo subtitles to news segments, reaching speakers who prefer the language over Spanish.
• Linguistic documentation: Field linguists can quickly transcribe hours of interviews, freeing time for analysis.
• Accessibility: Real-time captions for Mayo-language webinars and meetings ensure full participation.

How Speechyou Helps

Speechyou is the only AI transcription platform that explicitly supports Mayo. Our system is designed for low-resource languages, offering:

• High accuracy (>95%) on clear Mayo audio, including vowel length and glottal stops.
• Support for code-switching with Spanish, automatically recognizing both languages.
• Dialect adaptation: Users can upload samples of Sonoran, Sinaloan, or Mountain Mayo to improve results.
• Subtitle generation in SRT and VTT formats, ready for YouTube, Vimeo, or local media players.
• Unlimited transcription included in the Solo plan, making it affordable for community projects.

By providing accurate Mayo speech-to-text, Speechyou empowers speakers to preserve their language and share it with the world. Whether you are a linguist, educator, or community activist, our tool turns spoken Yoreme into written legacy.