Lingao Speech to Text: Preserving an Endangered Tai-Kadai Language with AI

Introduction

Lingao (Ong Be) is a Tai-Kadai language spoken by about 600,000 people on the island of Hainan, China. It belongs to the same language family as Zhuang, Thai, and Lao, but it has developed its own unique phonology and vocabulary over centuries of isolation. Today, Lingao is considered vulnerable, as younger generations increasingly use Mandarin Chinese in daily life. Accurate speech-to-text technology can play a vital role in documenting, preserving, and revitalizing the language.

Where is Lingao spoken?

Lingao is primarily spoken in Lingao County, on the northern coast of Hainan Island. The language is also present in parts of Chengmai County, Haikou, and other areas where Lingao people have migrated. There are three main dialect areas:

• Northern Lingao – around the county seat, most widely used
• Southern Lingao – farther south, with distinct vowel shifts
• Coastal Lingao – along the coast, influenced by Hainanese and Vietnamese

All three dialects are mutually intelligible to a degree, but differences in tone and vocabulary can cause confusion for automatic systems.

Why accurate speech-to-text for Lingao matters

Lingao lacks a strong written tradition. Most speakers are literate in Chinese but not in their own language. This means that oral traditions, folk songs, and everyday conversations are rarely captured in writing. With speech-to-text, we can:

• transcribe oral histories for archives
• generate subtitles for TV and social media
• create bilingual educational materials
• provide accessibility for the elderly and deaf

Without this technology, the language risks being lost as older speakers pass away.

Specific transcription challenges

Tonal complexity

Lingao has six tones: high, mid, low, high-falling, low-rising, and checked. Many words are distinguished only by tone. For example, maa (high) means “dog”, maa (low) means “horse”, and maa (mid) means “come”. A standard ASR model that ignores tone will produce many errors. Speechyou’s model uses pitch tracking and tonal context to separate these minimal pairs.

Lack of training data

Publicly available Lingao speech datasets are extremely small. Our team has compiled a custom corpus of 200 hours of audio from the Northern dialect, plus additional samples from Southern and Coastal varieties. Using transfer learning from related Tai-Kadai languages (like Zhuang), we have achieved high accuracy even with limited data.

Dialectal variation

A model trained only on Northern Lingao may struggle with Southern or Coastal speech. Speechyou offers separate models for Northern and Southern Lingao, and a generic model that can handle coastal speech with slightly lower accuracy. Users can select the dialect that matches their audio.

Use cases for Lingao transcription

Podcasts and radio

Local radio stations in Hainan broadcast in Lingao for older listeners. Automatic transcription allows them to provide written summaries and searchable archives.

Subtitles for video content

Lingao-language videos on platforms like Douyin can now have accurate subtitles, increasing engagement and reach. Export as SRT or VTT for easy upload.

Academic research

Linguists studying Tai-Kadai languages can transcribe fieldwork recordings in minutes instead of hours. This speeds up documentation and analysis.

Oral history preservation

Elderly speakers tell stories about traditional life, farming, and folklore. These recordings are transcribed and saved for future generations.

Accessibility

Deaf speakers of Lingao can read captions generated from speech, making meetings and events more inclusive.

Education

Teachers can create bilingual worksheets and readers, helping children learn to read and write in their heritage language.

How Speechyou helps

Speechyou is the only speech-to-text solution that supports Lingao with high accuracy. Our AI is trained on the language’s tonal system and dialectal variations. You can upload audio or video files, get a full transcription, and download subtitles in SRT or VTT format. The tool is available on web and mobile, with a simple interface designed for non-technical users.

With Speechyou, the Lingao community can preserve its language, share its stories, and ensure that future generations can still hear the voices of their ancestors.