Update on Chinese OCR

[Pengyou]

btw, happy birthday on your first decade! You have made learning Chinese not only easier for many people...but for some people you have made it possible, even though still not very easy

I just did a search on this forum for "Chinese OCR" over the past year and a half and I did not find anything. Most of what I did find was more than 3 years old, which is about a generation in the technology spectrum.

Has anyone purchased any Chinese OCR software in the past few months? I would appreciate any reviews you have. Please make sure you mention the kind of equipment you are using, i.e. which cpu, how much memory, etc.

Thanx in advance!

[mikelove]

As far as current Chinese OCR apps, any comment on their relative merits would be transparently self-serving but even I would say that there a lot more options out there now than there were 3 years ago - same basic concept, but a number of interesting mixes of UI / database / etc, including several promising free ones. Google and Microsoft both offer very solid free Chinese OCRs now on their respective mobile platforms, actually. (our own system has improved a good bit too - just added support for crosshairs on Android and about to add that also on iOS - but again it's basically the same concept now that it was back then)

As far as developments in technology, I think we're still about 1-2 years' worth of mobile CPU improvements away from the next big "leap" in live Chinese OCR, the point when mobile CPUs become fast enough that it's feasible to execute a modern text-detection algorithm like the Stroke Width Transform in real-time. Once that happens, we can start detecting / recognizing / translating all of the textual content in an image at once (even a complex image like a street sign, something most OCR systems have trouble with now); this will make it possible to solve most of the biggest problems in current live Chinese OCR systems, making them both more accurate and more ergonomic.

But I think the long-term future of Chinese OCR rests with giant companies like Google rather than little ones like Pleco - their R&D budgets will enable them to do things that we can only dream of. Chinese learners might still find an OCR system like ours (however limited) useful because it integrates with our numerous dictionaries / flashcards / etc (something you're unlikely to see Google take much interest in), but the vast majority of future Chinese OCR users (including those poor unfortunate souls who don't know / aren't interested in learning Chinese) will probably be better served by free apps from bigger companies - the inevitable Google Translate OCR app for Google Glass is going to take the world by storm.

[Pengyou]

Thanks...that is a lot to swallow. I am not so interested in mobile apps but on something that I can use with my scanner. I have a copy of ReadIris Pro 12 that came with my HP laserjet MFP. I have not tried it yet because I read from older reviews that it was so inconsistent that it was more trouble than it was worth...but that was a few years ago.

[mikelove]

Ah, sorry for the confusion.

I believe the trend in desktop OCR lately has been to start doing everything on the mystical "cloud" - Google Docs for example has built-in Chinese OCR now, and while I haven't investigated this I'm sure a few of the big China-based cloud drive services do as well (or are working on it anyway).

For offline software, we're partners with Hanvon on a couple of projects, but we're partners with them for a reason - they make very good recognition algorithms - so their scanners and OCR software are well worth checking out.

[markcarter]

For anyone reading this specifically for mobile OCR software, here is a good roundup of the current offerings:

http://www.theworldofchinese.com/2013/04/ocr-app-faceoff-can-your-phone-read-chinese/

which includes our very own Hanping Chinese Camera.

Almost all the offerings use either a box or crosshair to specify which part of the image you want to read. However, we use a line similar to how a barcode scanner would work.