iListen 1.8, by MacSpeech
Posted: 8-Mar-2008

2 1/2 out of 5 Mice

Vendor: MacSpeech Type: COMMERCIAL

Reviewer: Lawrence Grant Class: PRODUCTIVITY

Overview
iListen 1.8 from MacSpeech is software that creates text from ordinary speech and interprets voice commands to be executed by the computer. It allows the user to enter text into most applications (such as word processors, email clients, etc.), navigate documents, and control the computer (i.e., open and close applications, execute functions within applications, etc.), without the keyboard or mouse, using the individual's ordinary voice for text, and simple vocal commands to control the computer and its applications.

The software is intended for essentially two kinds of people: those who want to increase their productivity (speech is considerably faster than typing for creating text, and for those who already dictate, the software can eliminate intermediate transcription functions), or those who have difficulty using keyboard and mouse.

iListen is commercial software that can be purchased directly from MacSpeech, at Apple Retail locations, or many other online outlets such as Fry's, Amazon, etc.. A complete listing of retail sources can be found at the
MacSpeech web site. The basic software with a headset and noise-canceling microphone costs approximately $135. The software is unique in that it is designed specifically for the Mac.

System Requirements

  • Mac OS X 10.4.0 or later (version 1.8 is designed for 10.5.1 or later (Leopard).
  • 512MB (or more) of RAM.
  • A Mac that shipped with a G4, G5, or Intel processor.
  • 70MB of Free Disk Space, plus 20MB for each voice profile created.
  • A MacSpeech-certified USB noise-canceling microphone is strongly recommended.


For this review, the computer is a MacBook Pro, 2.4 GHz with 2 GB of RAM running OS X 10.5.1.

Installation
The basic installation of the software is straightforward. Insert the CD, mount the disk image, and simply drag the application icon to the alias provided for the OS X Applications directory.

In Use
Using the software, however, requires fairly complicated setup and training procedures. The Tutorial document, while giving a good overview of the program, skims over some steps. The User Manual, on the other hand, is detailed, and step-by-step. Even if you are a user who is accustomed to "plugging in" the software and using minimal trial and error to make it work, you'll struggle with the software with referencing the manual.

On the first launch of the program, iListen will bring up a Profiles dialog box. You will need to create a voice profile by clicking the "New" button. A second dialog box will ask you to name the profile, and ask you to select whether you are male or female.


iListen Profiles


It's important to know that iListen does not work by translating your words, word-by-word, into text. It works by using a sophisticated algorithm to translate your voice into text. Even though you are "saying words," the program is translating your unique "sounds" into word text, or into specific commands to control the Mac. The voice profile is stored as <name>.voice in your Documents directory.

The next step after the voice profile is created is to set up your microphone. There are explicit recommendations for positioning the microphone that are best to heed. If your microphone is properly plugged into a USB port, a dialog box appears prompting for the recording volume. You will read a bit of text until it disappears, indicating that the audio level from your microphone has been properly set. Afterwards, a Recording Quality panel will play back what was recorded so that you can check for "odd" sounds that might indicate a faulty microphone connection.


Recording Volume


Finally, you will be asked to speak a series of numbers as they appear in a Recognition and Silence Detection panel. This is how iListen learns to interpret the "silence" in your environment.

At each stage of this microphone setup, the Users Manual gives you fairly detailed troubleshooting directions in case the process doesn't go as expected. I used the setup procedure on 3 different machines without any problems.

After the microphone setup, iListen moves to the "Learn My Voice" training to "personalize" your voice profile.


Voice Training


For some, the training can take as little as five minutes. For most people, the training will be a major time investment. Admittedly, I found the training fun, albeit time consuming. First, you are presented with a series of 6 panels. Each asks you to speak the material in the panel, being sure to speak all the punctuation, such as COMMA, PERIOD, PARAGRAPH, and the like. The program tracks along with your reading, turning the material green as it recognizes what you are saying. A yellow arrow keeps track of your position in the text.

As you finish each panel, a new one appears, until you have read through all six. If you make a mistake, or iListen doesn't "understand" what you said, the yellow arrow will stop and turn into a red question mark. Continue reading from that position in the text.


Retraining


When you have finished this "Welcome" story, you can "Finish Training," or select another story to read.


Training Completed


Generally, more reading results in more accuracy, with some exceptions. For example, if you try to read and train too much in one session, your voice will fatigue, which actually degrades accuracy. Once you've achieved 90 percent recognized text, more reading will likely do more harm than good (according to the manual). There are further training approaches that will help to improve accuracy.

Once you have finished this initial training, the program takes roughly as much time to analyze your voice as it took for you to read it. It's difficult to tell that this is going on. Without an obvious indication of something processing, it leaves you with the sense that the program may have locked up. It hasn't - you just have to give it some time to process what it has "heard."

There are two other "learning" modes for iListen. The second learning mode is called "Learn My Writing Style," and the third one is "Corrections" (making corrections in text you have dictated).

In the "Learn My Writing Style" mode, the user provides one or more documents that he or she has written for the program to analyze new words and adjust to individual speech patterns. The program will use these documents to add words to its "vocabulary."


Learning Writing Style


The third learning mode is available after you have dictated some material. This is called the "Corrections" mode. The "corrections" part of the software was probably the most difficult to get a handle on. The Tutorial was not enough to learn this mode, so I had to revert to the User Manual to figure out how to use this mode (or so I thought). When I was running the software using OS X 10.4.11, this process was supposed to work as follows: Dictate a paragraph or two, and if it is rather badly translated, switch to the corrections mode. Mark out the errors in the text area of the corrections screen with the mouse, have iListen play back what you really said, and if you said what you think you said, then choose or type in the correct words. At the end of the corrections, give the command "commit corrections," specify which window to which you wish the corrections to apply, then move on.

Unfortunately, using the version of iListen (v1.8) that was updated to work with Leopard (OS X 10.5), the text document was not only not corrected, it became hopelessly scrambled. The only solution would be to make a copy of the text needed, and work with that copy for the corrections training. Of course, that would not make any changes in the original text, but it would save it from being hopelessly scrambled.

To use iListen for dictation, the corrections mode really needs to work. When the transcription process makes an error, the user needs to know what he or she really said in order to both edit the document, or correct the behavior of the program.

For example, the following two paragraphs are a rough draft for part of this review.

Original Dictation:

The basic installation of the software is straightforward. Insert the CD, mount the disk image, and simply drag the application icon to the alias Applications icon. Of course, if you want to install the software in another location, you would then have to drag the application folder from Applications directory to the directory of your choosing. How (or if) locating the software in another directory will work is beyond the scope of this review. Using the software, however, requires fairly complicated setup and training procedures. I recommend that the user rely on the User Manual rather than the Tutorial document. The latter skims over some steps that I found confusing, while the manual is detailed, and step-by-step.


Here is what resulted from my dictation of this text:

The basic installation of the software is Street for work. Insert the CD, Mount the disk image, has simply drag the application icon onto the alias applications icon. Of course, if you want to install the software in another location, he would then have to drag the application folder from applications directory to the directory of your choosing. How all (or deaf) locating the software in another directory will work is beyond the scope of this review. Using the software, however, requires fairly complicated setup and training procedures. I recommend that the user rely on the user manual rather than the tutorial document. The latter skims over some steps that I found confusing, all the manual is detailed, and step-by-step.


The following is the result of the corrections process:

straightforward. Insert theThe basic installation of and simplye is Street for work. Insert the CD, Mount the disk image, has simply drag the application icon onto the alias applications icon you wouldse, if you want to install the software in another location, he would then have to drag the application How (or ifcations directory to the directory of your choosing. How all (or deaf) locating the software in another directory will work is beyond the scope of this review.Using the software, however, requires fairly complicated setup and training proceedures. I recommend that the user rely on the user manual rather than the tutori whileocument. The latter skims over some steps that I found confusing, all the manual isplugging in" the software


As you can see, the basic transcription is not that bad. Therefore, if you can remember what you dictated, editing the material would be fairly simple and quick. In the Leopard version of the software, however, the "corrections" process is badly broken.

I had actually hoped to use iListen 1.8 to write this review, but in several parts of the review where I wanted to discuss the use of commands, the program executed the commands, themselves, rather than recording them. The result was a garbled review. There are means to prevent this from happening within iListen, but it seemed more time consuming to get it trained correctly, leaving me with doubt about an increase in productivity.

Once the program is set up, and the initial training is complete, the primary interface consists of two panels. One panel displays several different program components, and the location for selecting those components: Whether or not the microphone is on, what mode is currently operational (dictate or command), what the program "heard," and correction mode selection.


iListen Panel


The second primary panel is a list of commands that are available to the user at that time.


Dictation Commands


The program does a pretty good job of turning speech to text. I was disappointed, however, with the "Corrections" and "Learn My Writing Style" modes. For example, in the transcription example presented earlier, I used the word "straightforward." In my experiments with the program, I corrected that word at least three separate times, and each time, when I used the word in a sentence, it was transcribed as "Street for work."

With respect to the other main mode of operation - the command mode - while I didn't attempt an exhaustive test of every command available (there are hundreds), they executed flawlessly. So if you are mainly interested in using iListen to control your computer and application functions, you may well be pleased with this version of the program.

As stated earlier, iListen 1.8 comes with a Tutorial manual, and a more extensive User Manual, with the User Manual being far more detailed. It also comes with a noise-canceling headset, a microphone and a USB adapter. The headset is functional, but not of high quality, and the microphone and adapter are a bit clunky. That said, it worked as expected.

The support for the program at this time is weak. When I first started to examine iListen a couple of months ago, I was able to use the web site and support easily. In the last week, it has changed substantially, and has become more difficult to use.

The big news, however, is that MacSpeech announced at MacWorld that they have licensed the Dragon engine for speech recognition. This engine from the Windows world has received great reviews in its "Naturally Speaking" form for Windows. MacSpeech has announced that the Dragon engine will be the basis of a new program currently named "Dictate." My assumption is that the vendor has found the engine underlying iListen to be deficient, and they are changing horses.

Summary
iListen 1.8 from MacSpeech is software that creates text from ordinary speech and interprets voice commands to be executed by the computer. It allows the user to enter text into most applications (such as word processors, email clients, etc.), navigate documents, and control the computer (i.e., open and close applications, execute functions within applications, etc.), without the keyboard or mouse, using the individual's ordinary voice for text, and simple vocal commands to control the computer and its applications. The setup and training required is quite extensive, and the results (at least for dictation) can be spotty. Since the vendor, MacSpeech, is in the process of revamping the software to use an entirely new voice transcription engine, I would recommend that potential users wait to see how this new program performs. The name of the program is "Dictate", and should be available by the time this review is posted. iListen 1.8 has great potential, but there are serious problems with the Leopard version. Hopefully these problems will not carry over to the next iteration of the software.

Pros

  • Command mode works extremely well
  • Microphone functions as expected
  • Does a fair job of transcribing simple dictation

Cons

  • Expensive
  • Correcting for improved accuracy is broken
  • Overall accuracy of transcription is highly variable from speaker to speaker
  • Current version is based on an "engine" that is being replaced


Overall Rating

2 1/2 out of 5 Mice