App Review: Predictable 4 – Part 2

by Cyndee Williams Bowen, MS, CCC-SLP

More than a year has passed since I wrote App Review: Predictable – Part 1 to publish my impressions of Version 4 of the Predictable Augmentative and Alternative Communication (AAC) application by Therapy Box. I promised to write a follow up blog when I completed the process of building my customized voice using a Beta version of ModelTalker speech synthesis software by Nemours Speech Research Laboratory.

Business got busier, life got hectic, my ancient laptop got squirrely, and it took much longer than anticipated to finish recording the 1600 sentences required to generate my unique electronic voice. I completed the last of the speech samples a few weeks ago, and in less than 24 hours I received ModelTalker Cyndee!

ModelTalker Cyndee

I’m now ready to share my electronic voice with you via voice recordings of the familiar Grandfather Passage. Please listen to all three samples before reading my analysis and conclusions.

Voice Recording 1: Cyndee reading The Grandfather Passage     

Voice Recording 1 is my natural voice and is therefore the target I hope to approach with ModelTalker Cyndee.   

 

Voice Recording 2: ModelTalker Cyndee reading The Grandfather Passage – Original

ModelTalker Cyndee reading an unedited version of The Grandfather Passage.   

 

Voice Recording 3: ModelTalker Cyndee reading The Grandfather Passage – Revised

ModelTalker Cyndee reading an edited version of The Grandfather Passage. Please listen carefully to see if you can identify changes I made.

 

Analysis

Since Voice Recording 1 is my natural voice, it is what it is.  I have no known vocal pathology and captured this recording with the same microphone that I used while recording all 1600 samples.

Voice Recording 2 is ModelTalker Cyndee reading The Grandfather Passage as published in The Source for Dysarthria ©1997 LinguiSystems. This recording presented a few problems. The most obvious is that electronic voices sound robotic. I identified and made several edits that I thought might enhance the end result.

Voice Recording 3 is ModelTalker Cyndee reading an edited version of The Grandfather Passage. I employed creative license to make a total of 15 edits targeting improved prosody, decreased co-articulation conflicts, and correction of one glaring vowel error where ModelTalker Cyndee pronounced the word banana as  bee-næ-nuh. I suspect the traces of my Southeastern USA dialect resulted in my pronunciation of the word as bih-næ-nuh and ModelTalker Cyndee interpreted the ih as ee.

I had to employ creative solutions to correct a couple of pronunciation errors. Specifically, in the first sentence, ModelTalker Cyndee pronounces the word know as known. I tried numerous spellings and punctuation to improve the end result. Adding a comma after the word eliminated the final n.

Commas were also used to facilitate articulation of word-final consonants that are often dropped in SE USA dialect. Adding commas improved the phrase old black frock coat and the word pronounced remarkably. And the word or in snow or ice was highly robotic and distracting. A comma placed after snow helped tremendously.

Finally, I took creative license with the text itself, changing frock coat, usually several… to frock coat with several. There were too many bits when talking about Grandfather’s voice, so I changed the second one to little. And …open air each day was cumbersome and the word air was difficult to understand, so I changed the sentence to …takes a short daily walk in the open-air, deleting each day entirely.

 Conclusions

All in all I am very happy with ModelTalker Cyndee! I think she sounds very much like me, and I look forward to exploring ways to use her with Predictable 4 to create email, text, and other messages. The ModelTalker program is still in Beta testing, so it will be interesting to see how it improves as the researchers at Nemours work out the kinks!

In the meantime, I’ve found that adding punctuation in odd places, changing vowels, and editing text to decrease negative effects of co-articulation are tools I can use to help my clients customize their AAC applications to better represent their natural vocal identity.

So, how about some input? What do you think? Does ModelTalker Cyndee approach the sound of my natural voice?

Bowen-Logo_Web-Emblem-EditedSmall

Disclaimer: I have no personal or financial interest or connection to any person or business entity related to this product.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s