question about different OCR (optical character recogni) programs

[jimdagys]

I have been experimenting with different ways to extract the regular subtitles from DVD into srt files. I used Subrip and
Subresync to do the OCR. (Both of these programs were suggested to do OCR). Both programs required typing for about 6 minutes of letters and symbols. The final srt file sometimes was good and other times had quite a few mis-spellings. Then I tried DVD Subedit. The OCR in that program required no typing. The entire DVD was extracted into srt file in about 8 seconds. And there were only a few spelling mistakes. My question is, why does DVD Subedit do such a better job than the other programs? And why would anyone even bother with programs like Subrip and Subresync to do OCR?
One other question. One of the spelling mistakes I found is a capital N was always lY.
Does anyone know a simple program that can change and correct all instances of such a mis-spelling?

[Baldrick]

It uses http://jocr.sourceforge.net/ for OCR and it's seems to be better than the others...but I guess the dvdsubedit creator(jeanl) will answer this better.

Use a text editor to replace lY with N and be sure to use a case sensitive replace like in ultra edit or you will replace every *ly* also.

[jimdagys]

Thanks for the info about UltraEdit. It works well to get rid of those kinds of spelling mistakes.

[jeanl]

Thanks for the info about UltraEdit. It works well to get rid of those kinds of spelling mistakes.

It works very well in most cases, but sometimes it does have problems with certain fonts. The reason why it works so well is that the creator must have spent hours upon hours meticulously coding and testing his recognition engine, which is entirely rule-based (no machine learning, it's all hand-code, if you can believe it).
I have made small improvements to it (modified some of the rules to get rid of some of the errors, and also added code to deal with italics - the original code didn't do well with italics).
But kudos to Joerg Schulenburg for his amazing work...
Jeanl