RootsTech 2015

Some people eat, sleep and chew gum, I do genealogy and write...

Sunday, November 24, 2013

More about voice recognition and genealogy

For the first time I have been using voice recognition software consistently the past couple of weeks. The accuracy is quite good but you have to be very careful to proofread the text produced. Although the program, Dragon Dictate, does a very good job of transcribing my voice, it does drop words occasionally and substitute words for unknown terms. The key to extensive recognition is to provide more than the minimal amount of training and allow the program to copy a substantial number of text documents that you have already produced. This helps to build the program's vocabulary and thereby increase its accuracy.

The main question is the program's utility directly involving genealogical activities. I think using the program to fill out forms such as entering names into pedigree charts or family group records would be frustrating due to the variety of names. You can use the spell function or the program will spell out each word letter by letter but at that point there is no advantage over entering the text by keyboard unless you are unable to use a keyboard. Either way, you need to be more consistent and proofreading anything that is produced.

During the past couple of weeks, if you have been reading my blog, you may have noticed some strange typographical errors. These can be attributed mainly to the use of voice recognition and the difficulty of spotting those errors even when you are carefully proofreading. But then again, you may have been noticing strange typographical errors all along none of which could be attributed to voice recognition. When the program substitutes the wrong word into a phrase or sentence, it is sometimes very difficult to pick that up due to the fact that the word is very close to the one intended. For example, it may drop one single letter thereby changing a word such as "none" to "one."

I must admit, that generally I do a rather poor job of proofreading. Perhaps, you can attribute this to the quantity of writing that I do, sometimes under pressure to finish a job quickly.

One of my commentators called my attention to the fact that Amazon.com has Dragon NaturallySpeaking home edition 12.0 at essentially half-price. In the software industry, this generally means that the developer, in this case the nuance.com, is getting ready to release an upgrade to the program. But it may also mean that they simply need to make sales.

One of the most difficult adjustments to using voice recognition software is changing the way that you write to reflect your method of dictation. Because of my work experience, I had used dictation extensively for years until computers became easier to use. The process of dictation was cumbersome because I had to dictate the document to the secretary then proofread the dictation after it was transcribed. Then the document had to be typed again with the corrections. When I began typing all my own documents, my secretary (now called legal assistants) would simply make corrections to my typed copy, thereby saving the time necessary to redo the entire document.

If you are using voice recognition software, you can watch the screen carefully and catch most of the errors that need to be corrected as you dictate. I have tried using voice recognition to transcribe handwritten documents such as journals or other source material. But because of the need to transcribe exactly what was written this may or may not work. When transcribing an historical document, it is necessary to transcribe even spelling errors and punctuation errors. Obviously, the voice recognition program spells the words correctly and therefore using voice recognition becomes more of a problem than a help.

In summary, I think that voice recognition gives a distinct advantage in the speed of entering text. It appears to me that I can dictate a blog post in about half the time that it would normally take to keyboard the same document. Depending on your accuracy and typing, you may find that voice recognition is either better or worse than your typing skills. In my case, I think it is about the same but I do make different typographical errors while typing then the program makes during dictation. One thing that I can say, is that voice recognition is finally reached the point where it can be used on a routine basis without driving me crazy.

This entire post was entered using voice recognition. I proofread the text as it was dictated and then went back and reread the document to see if I could catch any further errors. So, you can tell from this document the quality of the text that can come from the careful use of voice recognition. If there are any typographical errors, which I assume there are, they are probably the same kind of errors I would have made by keyboarding the text.

I realize to some extent this post covers points I have made in previous posts, some recently, but because of my intense use of voice-recognition software this past week I thought it was a good idea to update my thoughts on the process.

6 comments:

  1. You would think that a voice recognition program should at least be able to properly interpret the word "voice." See the first sentence of your blog. ;-)

    ReplyDelete
    Replies
    1. Thanks for one more edit. I guess that proves my point.

      Delete
  2. Reading your opening line gave us a clue you might be using voice recognition software. Check the word right before "recognition"

    "For the first time I have been using force recognition software ..."

    Thanks for the chuckle!

    ReplyDelete
    Replies
    1. Thanks. As I said above, I guess it was good illustration of the word substitution problem and the difficulty of proofing the results.

      Delete
  3. I see *then* for than, next to the last paragraph. It is difficult to see things like this, I think your eyes see what we believe it should be.

    ReplyDelete