Pages

Wednesday, April 3, 2019

When will we have handwriting recognition?


Automated, computer based, handwriting recognition is the dream of every genealogist. The challenge of handwriting recognition goes way beyond the now common Optical Character Recognition software that can recognize typed or typeset text, There are no standard sets of characters to work from; each person's handwriting is distinct and personal. It is true that the letters adhere to an underlying pattern, but the variations are overwhelmingly complex. Fortunately, there are a lot of developers around the world that are working on the problem.

One of the leaders in handwriting recognition development is the Brigham Young University Family History Technology Lab, on the campus of Brigham Young University in Provo, Utah.

https://fhtl.byu.edu/
 To see a good overview of the Lab's activities, including their efforts to develop handwriting recognition, see the slides from their RootsTech 2019 presentation.

Progress in this area parallels the increased memory and processing capacity of computers as well as the increased availability of high quality digitized documents.

There are two facets to the present investigations into handwriting recognition: those efforts that are aimed at the recognition of handwritten notes on tablets and those that attempt to recognize paper-based handwriting. Most of the efforts are aimed at present commercial applications and not on reading the handwriting in old historical documents. There are thousands of academic articles online on the subject.

After watching this development for years, I have seen incremental increases in both recognition and accuracy, but the dream of having all of the old handwriten documents transformed into text still appears in the distant future.

2 comments:

  1. Not that distant, see https://read.transkribus.eu/

    ReplyDelete
    Replies
    1. Thanks for the link. In international competition for handwriting recognition, BYU has received top honors. They are at 98% recognition for old German script. But there is still a long way to go.

      Delete