.

Terry Flanagan: With Today's Technology, Tolstoy Could Have Written 'War and Cheese'

Speech recognition software leaves me speechless.

I’ve often wondered how the great writers of the past would have fared with the authoring and publishing tools available today. Had Dickens been able to use Microsoft Word on a computer and had wireless Internet available in his hotel room during his many speaking tours, would he have been even more prolific than James Patterson? What would a Dickens website or Facebook page look like? What would Dickens tweet to his followers? Would he be blogging and responding to comments from readers? Would he be getting hate e-mail for killing off Little Nell or requests to do a sequel to Oliver Twist?

There’s no question that technology and the web would have provided Dickens the capability to be far more productive. Instead of a couple of dozen books, Dickens might have written dozens more. But he might have frittered away his time playing Pogo or Farmville instead. We’ll never know.

I can say that that the one technology that probably would not have helped much is speech recognition or speech rec as it is often called. Since my typing speed seems to top out at about 40 words per hour, despite years of having been at it, I recently started looking at alternative input methods. Given my experience trying to find decent transcription software, I think there’s a good chance that Dickens, or anyone else for that matter, might find speech recognition software as frustrating as I have.

It’s a bit like dictating to a complete idiot, who not only doesn't understand a single word you’re saying, but doesn't have the wits to string together a single comprehensible sentence.

After seeing the word warranty interpreted as “war and cheese," I expressed some doubts about the intelligence of the software to a friend of mine who is using the same program. In what context does the phrase “war and cheese” even make sense, unless there is some remote possibility that Tolstoy’s pet mouse also wrote an epic novel. He said that I needed to enunciate. Then we took a recording he made and passed it through the transcriber. Upon reading the text output, he commented that it needed a little work. I agreed, unless his intent was to produce utter nonsense that bore little semblance to the original document. I wonder what kind of grade that essay would have gotten in English Lit.

I am convinced that if writers had to rely on speech recognition as it exists today, we would have bookshelves full of complete gibberish that would take more than the Rosetta Stone to decipher. That’s not to say that some of it would not be entertaining. Mangling the language has been used for humor from Goldsmith’s Mrs. Malaprop to comedian Norm Crosby. But speech recognition software elevates that art to a new level.

My friend suggested I try using an iPhone app that he likes which incorporates speech recognition. I tried the word “coincidence” and was given the three choices of “at, At, or AT.” I’m not even sure what the distinction between the choices is in this case. Apparently four syllables were too much for the program to analyze, and it began stuttering.

I then watched an online demo of someone using another speech recognition application and, of course, the speaker was successful at getting it to recognize what he said. This looks simple, or so I thought. I tried playing a short recorded message into the same application and didn’t come close to getting a single word right. It was almost like a random word generator. I next tried the same phrase spoken in the demo and managed to get one word right. Not exactly encouraging.

Obviously, there is a ways to go in computer speech recognition. There has been some success with limited vocabularies, such as those you may encounter in telephone IVR applications. Speaker training also helps somewhat, but even systems that use speaker-dependent recognition have problems recognizing speech unless it is spoken in slow and deliberate monotone, not at all the way people speak normally. Background noise and conversations are also impediments to speech recognition. The computer is simply no match for the human brain when it comes to processing and recognizing speech.

Having worked at a company in the '80s that was working on speech recognition, I know that a lot of geniuses are working on the problem and will eventually solve it. FYI, I was not working on that project or we might have produced a machine with a speech impediment and a hearing impairment. Speech recognition and speech synthesis have certainly improved over the years, though. I expect that one day we’ll be fully conversant with our computers and mobile devices. Our machines will then probably be correcting our speech, secretly snickering at us behind our backs, and taking and posting embarrassing pictures of us on Facebook.

This post is contributed by a community member. The views expressed in this blog are those of the author and do not necessarily reflect those of Patch Media Corporation. Everyone is welcome to submit a post to Patch. If you'd like to post a blog, go here to get started.

Boards

More »
Got a question? Something on your mind? Talk to your community, directly.
Note Article
Just a short thought to get the word out quickly about anything in your neighborhood.
Share something with your neighbors.What's on your mind?What's on your mind?Make an announcement, speak your mind, or sell somethingPost something
See more »