Turkish Language Support & Confidence Score

[AlperenEvci]

Hi,

I have been exploring PaddleOCR V5 and its multilingual text recognition capabilities. While I found that many languages are supported (including Latin-script models), I could not find any explicit mention of Turkish support in the documentation.

My questions are:

1. Is there currently official support for Turkish (including special characters such as “ç, ğ, ı, ö, ş, ü”)?

2. If not, are there any plans to add Turkish language support in the upcoming versions?

3. Does the recognition API/model return a confidence score along with the predicted text (for Turkish or other languages)?

If Turkish is not yet supported, I would be interested in contributing by preparing a Turkish character dictionary and training dataset. Could you please share some guidelines or best practices on how to add a new language properly and contribute it back to the repository?

Thanks in advance for your guidance!

[liuhongen1234567]

Hello, the current version of the model does not consider certain special Turkish characters. These special characters such as “ç, ğ, ı, ö, ş, ü are planned to be supported in PaddleOCR 3.3.
At the same time, for the recognition module, the confidence score can be obtained through res['rec_score']. The sample code is as follows:

from paddleocr import TextRecognition
model = TextRecognition(model_name="PP-OCRv5_server_rec")
output = model.predict(input="general_ocr_rec_001.png", batch_size=1)
for res in output:
    print(res['rec_score'])
    res.print()
    res.save_to_img(save_path="./output/")
    res.save_to_json(save_path="./output/res.json")

For more parameters, you can refer to the text recognition module documentation: https://www.paddleocr.ai/main/en/version3.x/module_usage/text_recognition.html#3-quick-start

[liuhongen1234567]

For contributions and collaborations regarding the Turkish character dictionary and training dataset, feel free to send an email to paddleocr@baidu.com.

[CapturefastCEO]

The emails sent to paddleocr@baidu.com are bouncing back.
I have millions of Turkish documents. I am eager to help.