[KongchangAI]
Modelufakzeka-kulak

ufakzeka-kulak

A lightweight open-source Turkish speech recognition model based on a FastConformer encoder and TDT/CTC decoders, supporting output with punctuation and capitalization.

Timeline (last 90 days)

Nov 3

ufakzeka-kulak is an open-source Turkish speech-to-text model with only 31.6M parameters, using a FastConformer encoder and TDT and CTC heads, trained from scratch, and released under the Apache-2.0 license.

Unverified50%
Nov 3

The model outputs Turkish text with capitalization and punctuation, with numbers presented as words.

Unverified50%
Nov 3

The model used 4,977 hours of training data, all sources of which permit commercial use, of which 281 hours are human-transcribed and the training text for 4,696 hours comes from uploader subtitles or machine transcription.

Unverified50%
Nov 3

For the 4,696 hours of non-human-transcribed data, a sample is retained only when a second independent reading of the audio matches the text.

Unverified50%
Nov 3

With greedy TDT decoding, the model achieves a normalized WER of 0.0921 on the FLEURS Turkish test set, with a 95% confidence interval of 0.0792 to 0.1055.

Unverified50%
Nov 3

On both the FLEURS and Common Voice datasets, ufakzeka-kulak's WER is lower than Whisper base and Whisper small, and the paired confidence intervals are entirely below zero, so the performance advantage is statistically significant.

Unverified50%
Nov 3

In conversational speech scenarios, ufakzeka-kulak's WER rises to 0.3807.

Unverified50%

All Facts (7)

Source Articles