Google has offered free optical character recognition (OCR) on images and PDF files uploaded to Google Docs for some time now, but is it any good?
I tested the feature out on different file types and had varying degrees of success. First, here’s how you use the feature:
Upload your files (images of PDFs) to Google Drive.
From within Google Docs, right click on the image or PDF and choose ‘Open with Google Docs’.
Wait as Google converts your document. Depending on the file size this might take a few seconds, so be patient. The converted document should automatically open.
Here’s the process condensed into an animated gif:
If you converted an image then the first page of the new document will be your image and the second page will contain the OCR’d text. For PDFs the text is inserted after each individual page.
It’s impressive, but by no means perfect. The example in the main image above was the best results I was able to get. From what I can tell there is only one incorrect detection – 哼 was incorrectly detected as 啤. It’s a small mistake, but it still means that you need to check the documents after converting to ensure they are error free.
Here are some other examples from different sources.
Images seem to cause issues, but this seems to be a problem with all OCR software.
Here I photographed the page of a book and uploaded it without editing.
Even though the image is dark, the OCR detection was very good.
The detection of vertical text was also very impressive.
The about page for Optical Character Recognition in Google Drive does list some limitations.
OCR on PDF files is limited to the first 10 pages, so if you are converting a large PDF you may to want to split it up first.
The maximum file size for OCR is stated as 2MB, but I tried a 14MB PDF and it worked, though only the first ten pages were converted.
OCR does not seem to be available on mobile at all. I tried inserting images directly into Google Docs, using the Scan feature of Google docs, and trying to upload an image to Drive and then import into Google docs. None of these methods resulted in an OCR conversion of the image being created. So for now it seems the desktop is the only place to use this feature.
The results from Google Docs OCR were surprisingly good. I was impressed by the limited detection mistakes and vertical text recognition. Images on the pages caused some issues so this may be more suited to regular books, and not magazines or other graphical materials.
The fact that there are mistakes make it difficult to recommend for anyone other than intermediate/advanced students and Chinese teachers. If your Chinese is good enough that you can spot and correct mistakes, though, then this could be a great way to import and edit documents.
Right now this would be most useful to you if you have the specific requirement of extracting text out of an image. If you just want to read the text then there are easier ways in the form of mobile apps from Hanping and Pleco.
Have you used Google Docs OCR before? What kind of results did you get?