Kennedy on Tech

Speech recognition moves past the dream stage


For many years, the compelling drive toward speech recognition has dead-ended in unsatisfactory attempts to implement it. Gradually it dropped off the technology radar for most lawyers.

Are there now reasons to think that the time for speech recognition has finally come?

We were often told that when processors got faster and stronger, and storage got larger and faster, and more data about speech was learned and incorporated, things would be different. Speech recognition often seemed a case of "it sort of works now, but soon it will be really good."

As Moore's law marched relentlessly onward, the power of chips doubled every 18 months, data storage got bigger and faster, new algorithms were developed, and a new era of speech recognition slowly began to unfold.

Add to the mix the cloud, the ability to process huge amounts of data at server farms rather than on individual devices and large databases of speech data. Then throw in mobile apps and widely available broadband access to the Internet. Suddenly, the speech recognition world is starting to look different—and promising.

Data input now can be done by traditional keyboards, tablet and phone keyboards, fingers and thumbs, pens, swipes and smart key commands. Each works best in certain contexts and for certain tasks, but it's no longer one size fits all.

I suggest that speech is now ready to become one part of the mix.

I realized recently that I was getting used to speaking numbers and commands on customer support calls, asking my phone to search Google by voice command, and dictating instant messages and emails, often perfectly. Perhaps you have noticed the same thing.

SPECIFIC USES

Speech becomes an interesting alternative when you must input data hands-free (think of dialing a call from your car); when speech can produce a text, email or draft as fast and as well as your typing skills (especially on a mobile device); or when it can do quick searches or initiate commands. In other words, by not asking speech recognition to do everything and being frustrated when it does not do it all, we can focus on where speech really helps us and take advantage of the help it offers us right now.

Here are some experiments you can try:

Data entry on mobile devices. On the keyboard in many apps, you will probably see a small microphone icon near the space bar. Tap it and you can start dictating. The text you speak will appear as if you typed it, often with no errors.

Hands-free use. We know texting, typing or dialing while driving is not safe; it might be illegal in your location. Using voice commands for dialing and handling simple tasks, either in your car's system or on a mobile phone, might help you be both productive and safe.

Commands and triggers. Some apps or systems, especially systems in new cars, have prebuilt commands that can be triggered by speech, like saying "call home" and your phone will dial your home number. Some apps allow more complex or customized voice commands. For common complex tasks with several steps, using a simple voice command can be a big time-saver.

This article originally appeared in the June 2015 issue of the ABA Journal with this headline: “Straight Talk: Speech recognition moves past the dream stage.”

Dennis Kennedy is a St. Louis-based legal technology writer and information technology lawyer.



We welcome your comments, but please adhere to our comment policy and the ABA Code of Conduct.

Commenting is not available in this channel entry.