Itza' (Latin script) Speech to Text: A Complete Guide
Itza' Speech‑to‑Text: Preserving a Mayan Language with AI
Itza' (also known as Itzaj) is a member of the Yucatecan branch of the Mayan language family. It is spoken primarily in the towns of San José, San Benito, and Santa Elena in the department of Petén, Guatemala. With fewer than 1,000 native speakers, most of whom are elderly, Itza' is classified as severely endangered. The language is written using the Latin alphabet with additional diacritics to represent glottalized consonants and vowel length.
Why Accurate Speech‑to‑Text for Itza' Matters
As the last fluent speakers age, the urgency to document the language grows. Oral histories, traditional songs, and everyday conversations are the only records of many lexical items and grammatical structures. Manual transcription is slow and expensive, especially for a language with no large commercial demand. AI‑powered speech‑to‑text can accelerate this work by producing draft transcripts that linguists and community members can then review and correct.
Specific Transcription Challenges
- Ejective stops: Itza' has a full set of glottalized consonants (p', t', k', tz', ch'). These are produced with a simultaneous glottal closure and a burst of air, which can be confused with regular stops by generic ASR models.
- Vowel length: Short, long, and extra‑long vowels are phonemic. For example, k'ak' (fire) vs. k'aak' (hot) differ only in the length of the first vowel. Accurate timing detection is essential.
- Limited training data: There are no large public corpora of transcribed Itza' audio. Speechyou addresses this by using a base model trained on Yucatec Maya and then fine‑tuning with a small curated dataset of Itza' recordings.
Use Cases
- Podcast and radio subtitles: Community radio stations in Petén produce programs in Itza' about health, agriculture, and cultural events. Speechyou can generate SRT subtitles, making the content accessible to those who prefer reading or are learning the language.
- Academic research: Linguists recording elder speakers can upload audio and receive a text file in minutes, compared to hours of manual transcription.
- Revitalization materials: Bilingual books (Itza'‑Spanish) can be created from transcribed oral narratives, preserving stories for future generations.
- Accessibility: Deaf community members who are literate in Itza' can follow spoken content through subtitles, bridging communication gaps.
How Speechyou Helps
Speechyou is the only commercial speech‑to‑text tool that supports Itza' out of the box. It uses a custom acoustic model that recognizes glottalized consonants and vowel length with high accuracy. The platform supports audio and video uploads, and outputs transcriptions in plain text, Word, PDF, and subtitle formats (SRT, VTT, ASS). Users can edit the transcript in a built‑in editor, and the system improves over time as corrections are fed back.
For a language with so few speakers, every accurate transcription is a step toward preservation. Speechyou empowers the Itza' community to document their language at scale, ensuring that the voices of the elders are not lost.
