Skip to main content
WeeMed Logo

Dedicated to creating intelligent information systems for healthcare institutions through innovative technology, enhancing medical service quality

WeeMed Technologies Inc.

VAT: 00085209

Follow Us

Quick Navigation

  • About Us
  • What We're Building
  • Solutions
  • EAP Marketplace
  • Services
  • Technology
  • News
  • Careers
  • Contact Us

Legal & Corporate Info

Privacy PolicyTerms of ServiceSecurity PolicyESG Sustainability

Contact Us

Phone
+886-4-23016388
Email
service@weemed.ai
Address
20F-1, No. 186, Sec. 2, Taiwan Blvd., West Dist., Taichung City
© 2024-2026 WeeMed Technologies Inc. All rights reserved.
ESG SustainabilityPrivacy PolicyTerms of ServiceSecurity Policy
WeemedAI Logo
Home
About
Products
What We're Building
Services
Contact

Taigi speech recognition

Older adults speak Taigi,and machines don't understand it

This page is about the speech half: why the frontline needs a model that understands Taigi, which step MediaTek had already done, which step we filled in, and the one we tried and haven't got right yet. Three of the four files we have on Hugging Face come from this work.

Why it's needed

The older adult is talking. The care worker is looking down, typing.

A care worker sits beside an older adult at a community station, tablet in hand. Did you sleep well last night? Do you get up in the middle of the night? The older adult answers. The worker has to listen, translate the answer into whatever the form wants, and type — all at once, head down. The older adult ends up talking to the top of their head.

And the older adult is speaking Taigi. Often Mandarin and Taigi land in the same sentence: "my leg, it goes sng (aches)." The worker understands it perfectly. They just have to convert it to Mandarin before they can type it.

What we wanted was simple: let them look up, talk to the person properly, and have the record write itself. That takes a speech model that understands Taigi — and off-the-shelf speech recognition handles Mandarin well and Taigi badly. Older adults are not going to switch to Mandarin to suit our software.

Someone in Taiwan had already built a model that understands Taigi. What we filled in was "it doesn't fit on the computer that's actually in the room."

Which step is ours

MediaTek had already done the hard part

Breeze-ASR-26 was trained and open-sourced by MediaTek Research (Apache-2.0, derived from OpenAI Whisper large-v2), and it understands Taigi and Taiwanese accents. We did not train it — that needs saying plainly.

But it is 3.1 GB and wants a GPU to run smoothly. The six-year-old PC at a community station cannot hold it, and a care worker's phone certainly cannot. So we did the part that was left: shrink it, actually measure the speed and the memory, build three versions for three kinds of machine, and put them back online.

  • By design the model writes Taigi out in Mandarin Han characters, not Taigi orthography. It gives you the meaning, not the Taigi reading.
  • On clips under three seconds with little content, every Whisper-family runtime starts guessing. That is a property of the family, not of the quantization.

Three builds

One model, three kinds of machine

The only difference is what hardware they run on. All Apache-2.0 — commercial use and modification without asking us.

  • Breeze-ASR-26-ct2

    The fast one, if you have a server

    Size
    1.5 GB (int8), about 2.9 GB memory while running
    Speed
    RTF 0.21 — nearly five times faster than real time
    For
    Servers, machines with 8 GB+, GPU hosts
    Runtime
    faster-whisper / CTranslate2
    Hugging Face model card
  • Breeze-ASR-26-GGML

    Works on one old laptop — just drag the file in

    Size
    848 MB (Q4_0) / 1.1 GB (Q5_0)
    Memory
    about 1.85 GB; runs on a 4 GB machine
    Speed
    RTF 0.40 on a 4-thread CPU
    Runtime
    whisper.cpp / MacWhisper / superwhisper — no code needed
    Hugging Face model card
  • Breeze-ASR-26-ONNX

    Taigi recognition on a phone or in a browser

    Format
    int8 ONNX (one encoder file, one decoder file)
    Speed
    RTF 1.3 on a 4-thread CPU — slower than the other two
    For
    Android, iOS, browser (WASM)
    Runtime
    sherpa-onnx / onnxruntime
    Hugging Face model card

On speech longer than five seconds you cannot tell the compressed version from the original (about 3.6% character error against it). On a home visit a phone is often the only device there is, which is why the slowest build had to exist too.

The one we didn't publish

If it isn't good enough, it doesn't go up

We also tried shrinking it to 237 MB so it could run on a phone directly. There was real progress: on genuine long-form speech, the gap to the original model closed from 63% to 38.7%. But the original model itself is at 14%. Still too far, so we did not publish it.

One failure along the way is worth writing down. On the first attempt we pushed the small model too hard to imitate the big one's manner, and it got worse. Progress only came after we put the right answers back in charge and let imitation play a supporting role. We publish records like that too, because they help other people more than a nice-looking number does.

What we are stuck on is not training technique. It is that there is very little recorded Taigi. Every open source added together comes to a few dozen hours, mostly short sentences read off a script — a long way from how an older adult actually talks.

There is a more fundamental problem too. The faithful way to write spoken Taigi is Tâi-lô romanization; writing it in Mandarin Han characters throws the Taigi reading away. Written in Han characters, 斷腦筋 gives the reader no clue that it means a stroke. So what we are building now outputs Tâi-lô. It isn't ready. When it is, it will show up on this page.

What it does in the product

Turning sound into words is only the first step

What comes out of speech is still how the frontline talks. 皮蛇 is still 皮蛇; it has to map back to 帶狀皰疹 before it can be stored — and that is a different model's job. Only with both halves does a spoken sentence become a usable record.

  • The model that reads those words: IlhaEmbed
  • How records get organized into the system: Intake AI
  • What we've released, and why we open it

Someone in Taiwan had already built a model that understands Taigi. We made it fit on the computer that's actually in the room.

Go use them.

All three builds are on Hugging Face under Apache-2.0. Each model card carries the measurement method, the limits, and what it is suited for. The upstream model is MediaTek Research's Breeze-ASR-26 — they trained it; we quantized, measured, and converted it.

Open our Hugging Face orgPress & partnership contact