Hacker Public Radio

Your ideas, projects, opinions - podcasted.

New episodes Monday through Friday.


HPR1769: A Demonstration of Dictation Software on my Office Computer

Hosted by Jon Kulp on 2015-05-14 00:00:00
Download or Listen

Transcript Performed by Dragon Dictate [dumped "as is"]

Hi everybody! This is John Kulp In Lafayette, Louisiana. I am going to do a rather strange episode today. What I'm doing is demonstrating the dictation software that I use on the office computer that I have here at work. If you listen to my previous episodes, then you have heard me speak of the blather speech recognition program that I use on my Linux desktop, but as you may also remember, blather is not a dictation tool. Blather is a tool where you have to set up commands that will run other commands. In other words, you have to configure everything from scratch. I do have some capabilities for dictation on my Linux desktop, but they involve using the Google Web speech API and a special dictation box that I have set up, and these are not at all good for longform dictation. For serious dictation, such as writing letters and memos and other longform text, you really need a proper dictation tool. These are available built into the operating systems of Windows and Mac OS 10, but I normally use the Dragon naturally speaking software instead. I have found that it is more accurate and more powerful than the built-in versions that you can get on either Windows or Mac. That doesn't mean you shouldn't try out the built-in speech recognition on Windows and Mac, you definitely should, because I think you would be very impressed with him. I know for sure that the version on Windows learns from your voice and from the corrections that you make to the text that you were spoken, and eventually becomes very powerful in recognizing your speech. The biggest problem that I had with the Windows speech recognition was that it was a huge memory hog and frequently brought my system to a grinding halt. This is not good. Blather never does that, but then again bladder cannot take dictation. The latest system that I use for dictation is on a fairly recent Mac Mini running the nuance Dragon Dictate software. This is a very powerful dictation program that learns from your speech patterns and you can also add words to the vocabulary so that it will get them right when it hears them. This is especially important to do if you have frequently used unusual words, such as a name with an alternate spelling from what is normally in the program's dictionary. One of the great things about the Mac Dragon Dictate program, also, is its ability to do transcriptions of audio files. In fact the reason I am speaking this way is that I plan to use the transcription of this recording as the show notes verbatim without any corrections. The difficulty that most people have with dictation software at least initially is doing things like punctuation and capitalization. You have to remember to do these things or else your transcript will come out without any punctuation or capitalization, unless the words that you are speaking are known proper nouns. It also capitalizes automatically at the beginning of the sentences, so that if you use periods frequently then you will have capitalized words after those periods. You can see that I'm having trouble speaking this text in a fluent way, and this is one of the other difficulties that people have when initially using transcription software. It works best when you can express complete thoughts without pausing, because it learns from the context of your words. It has algorithms that calculate the possibility of one word or another based on the context, and so it is much better to speak entire sentences at one than it is to pause while trying to gather your thoughts. This is a major difference from trying to write at the keyboard, where it does not matter at all if you pause for seconds or even minutes while you think of what you want to write next. Anyhow, I highly recommend using some kind of dictation software if you suffer from repetitive strain injuries like I do. This will save you many thousands of keystrokes. Even if it's only using the speech recognition that's available on your phones over the web, that's better than nothing. The disadvantage of any of these services that have to send your recording over the web to get a transcription and then send it back into your device is that they will never learn your voice and your particular speech patterns. In order for that to work best, you really have to use a dedicated standalone speech recognition program that resides locally on your computer and saves your profile and learns from your speaking. Well, I guess that is about it for today, I hope you have enjoyed hearing this brief lesson on dictation. See you next time!

Links

Credits

------------------------------------------------------------

Comments



More Information...


Copyright Information

Unless otherwise stated, our shows are released under a Creative Commons Attribution-ShareAlike 3.0 Unported (CC BY-SA 3.0) license.

The HPR Website Design is released to the Public Domain.