No best char error rate

[Meet Yogi]

Hi all,

I trained tesseract 4.1.1 with around 28000 data in which it's training started with best error of 4.187 and I got best error of 0.057. Now I added further data and final training data is around 1lac images. But the training starts with worst error rate of around 274 and never goes below 100. From it if I reduce the data to around 1500 images i will again start getting best error rate. I'm want to figure why I'm getting such a high error rate on large data.

Thanks you in advance.

Thanks & regards,

Meet Yogi