As much humor as autocorrect's mistakes have bestowed on us, it would be nice if our "smart" phones understood the most embarrassing of grammar errors, changing our "its" to "it's" and correcting all those small, but incredibly frowned upon grammar crimes that we commit far too often. In addition to the standard word dictionary, some of us hurried texters would benefit from a grammar dictionary. It's not that we don't know the difference between "we're" and "were," but when texting it's hard to get all the way over to the character screen and sometimes we're rushing and things just come out wrong and who proof reads text messages, anyway? Sure, there's growing social acceptance of these mistakes, but we know some of you judge us by our typos.
Well, lucky for us less-than-careful texters, the developers over atNuance — the company that invented T9 (the predictive text used on dumbphones which guesses, for instance, whether you wanted D, E, or F when you press the 3) and now provides predictive typing software for most of the major phone makers other than Apple — are working on just such a technology. It's called "future context" and it will look at an entire sentence, or a sentence fragment, and if it thinks that, say, an "its" should be an "it's," a little underline will show up, prompting the texter to change it.
As of right now, smartphone texting programs work off of what is called an "Ngram" model of language, which uses an algorithm to figure out likely strings of words. "You just basically have a big huge dictionary that lists all of the words from highest frequency to lowest," Aaron Sheedy, vice president of the Tech Input and Mobile Group at Nuance, told The Atlantic Wire. "So you have a way to algorithmically go through the tap sequence — there's a very complex algorithm trying to figure out which word you're going for."
So, if you write "I pledge," for example, the software has a pretty good indication that "allegiance" might come next, especially if you type "al," so it will autocorrect to that. That type of program can also account for certain aspects of grammar. Like, if you type "I" it will know to use the singular of a verb, like "am" instead of "are."
But, it doesn't work for anything that needs more context like "its" versus "it's" "'Its' is heavily dependent on whats about to be written, not what's already written," added Sheedy. The computer can't know the right version of the word without the full sentence. "Once we've written 'it's going to be a great day' versus 'its my thing,' then we know and then we can correct it."
That capability is exactly what Sheedy and his team are hoping to accomplish with the future context software. Using the same Ngram model, Nuance hopes to broaden its capabilities to take entire sentences into consideration. "We're working on how to take the Ngram to look at the right and left of any given word," says Sheedy.
For all of you language snobs lamenting the end of "proper English" due to all the typing on the Internet, don't worry: autocorrect for grammar will continue to degrade the mother tongue. This new and improved phone dictionary will only suggest proper grammar, not automatically correct it. And, if it learns that you like to subvert the stodgy old rules of standard English, it will conform — just as it always has — adding your unique usages of "its" along with those l8rs and lols. "Learning your own personal style of language is, for us, more important than adhering to a prescriptive model of language," added Sheedy. Good, because some of us like it that way.
Photo by Intel Free Press via Flickr.
As much humor as autocorrect's mistakes have bestowed on us, it would be nice if our "smart" phones understood the most embarrassing of grammar errors, changing our "its" to "it's" and correcting all those small, but incredibly frowned upon grammar crimes that we commit far too often. In addition to the standard word dictionary, some of us hurried texters would benefit from a grammar dictionary. It's not that we don't know the difference between "we're" and "were," but when texting it's hard to get all the way over to the character screen and sometimes we're rushing and things just come out wrong and who proof reads text messages, anyway? Sure, there's growing social acceptance of these mistakes, but we know some of you judge us by our typos.
Well, lucky for us less-than-careful texters, the developers over atNuance — the company that invented T9 (the predictive text used on dumbphones which guesses, for instance, whether you wanted D, E, or F when you press the 3) and now provides predictive typing software for most of the major phone makers other than Apple — are working on just such a technology. It's called "future context" and it will look at an entire sentence, or a sentence fragment, and if it thinks that, say, an "its" should be an "it's," a little underline will show up, prompting the texter to change it.
As of right now, smartphone texting programs work off of what is called an "Ngram" model of language, which uses an algorithm to figure out likely strings of words. "You just basically have a big huge dictionary that lists all of the words from highest frequency to lowest," Aaron Sheedy, vice president of the Tech Input and Mobile Group at Nuance, told The Atlantic Wire. "So you have a way to algorithmically go through the tap sequence — there's a very complex algorithm trying to figure out which word you're going for."
So, if you write "I pledge," for example, the software has a pretty good indication that "allegiance" might come next, especially if you type "al," so it will autocorrect to that. That type of program can also account for certain aspects of grammar. Like, if you type "I" it will know to use the singular of a verb, like "am" instead of "are."
But, it doesn't work for anything that needs more context like "its" versus "it's" "'Its' is heavily dependent on whats about to be written, not what's already written," added Sheedy. The computer can't know the right version of the word without the full sentence. "Once we've written 'it's going to be a great day' versus 'its my thing,' then we know and then we can correct it."
That capability is exactly what Sheedy and his team are hoping to accomplish with the future context software. Using the same Ngram model, Nuance hopes to broaden its capabilities to take entire sentences into consideration. "We're working on how to take the Ngram to look at the right and left of any given word," says Sheedy.
For all of you language snobs lamenting the end of "proper English" due to all the typing on the Internet, don't worry: autocorrect for grammar will continue to degrade the mother tongue. This new and improved phone dictionary will only suggest proper grammar, not automatically correct it. And, if it learns that you like to subvert the stodgy old rules of standard English, it will conform — just as it always has — adding your unique usages of "its" along with those l8rs and lols. "Learning your own personal style of language is, for us, more important than adhering to a prescriptive model of language," added Sheedy. Good, because some of us like it that way.
Photo by Intel Free Press via Flickr.
No comments:
Post a Comment