 |
| Dragon Dictation Free App |
What it is: A rather well known application that translates speech to text. You speak into it, and it gives you a text version of what you said. I've noticed that sometimes it recognizes raised intonation as questions, but normally, I just say "question mark" or "comma" if I want those included in the text. It is quite good and amazingly accurate if you speak clearly with relatively little background noise. It lets you easily bring up a keyboard if anything needs to be corrected, and even has drop down choices for some common homophones (like if it wrote "do" and you select it, it may automatically suggest "due" instead). The application also makes up a database of names in your address-book to help recognize these in your speech and spell them right. You can then export the speech you dictated into a text message, an email, some of the social networking programs, or lets you copy it for pasting (like into a note or whatever else you choose).
How we can use it in Tx: In some ways, this application is a measure of clarity and intelligibility. Granted, the clarity and intelligibility for this app needs to be superior to what would be required in normal conversation, but that's not a bad thing for treatment purposes. A patient that is working on intelligibility, volume, or slowing down their speech can practice with this app. Whether the patient is working on word-, phrase- or sentence- level, they can practice saying a list of words or phrases to see how many of the dictated words were clear enough for the application to transcribe. Makes it pretty easy to quantify progress as well. It is also quite delightful how good this app is at transcribing speech, which helps build motivation. This is a good companion to the Bla | Bla | Bla sound reactive app which helps work on volume and timing (see my write-up of the Bla | Bla | Bla app from March 11, 2012).
Goals we can target in Tx with this app: Intelligibility, pacing (especially slowing speech down at sentence or conversation level), volume, phoneme accuracy... and well, most dysarthria related goals. My PD patients have found it challenging but fun.
Some specific examples:
1. For word-level intelligibility or phoneme accuracy goals, start with a list of common words that target the phonemes in question, or a range of phonemes. I suggest dividing your list up by syllable length (so have a list of monosyllabic words, disyllabic, and longer). Have pt pronounce each word, and quantify success by how many trials got transcribed accurately. You can have each word said 5 times, and find % that were transcribed correctly, or just do the full list with each word once, and take % of correct words from the whole list. The transcription errors produced by the app should also help highlight the problem areas, especially for monosyllabic words (if a word is longer and one or two phonemes are off target, the app may still figure out what word it is, but shorter words rely on less context and more accurate production to transcribe). As such, you could use this app for evaluation purposes (an informal look at what sounds may be toughest for a patient, if so indicated).
2. For sentence-level pacing goals, have pt read a sentence or paragraph into the app and check on % of the sentence that was correctly transcribed into text. The feedback to the patient is very quick, and it may be helpful to see a visual of just how much of one's speech is misunderstood due to fast speaking. Putting pauses between words will up accuracy very quickly, and that would also be a good visual for the patient.
Well, you get the picture. Patient speaks into the app, the app returns with a text of what it was able to transcribe from the pt's speech. Accuracy is easily quantifiable, and suggestions can be made to improve the accuracy. Strategies are discussed, practiced, and pt tries again. Hopefully, accuracy goes up.
