Can machines truly understand language?
There are a lot of subtleties in language. Regions of every country have their own dialects, sentence structure is different for different languages and each dialect can have multiple types of slang.
Raytheon BBN has been given the almost-impossible task of developing a device that can perform two-way speech-to-speech translations, among other things. Now, I don't doubt a comprehensive translator can come from this, I doubt that any device can effectively translate human communication.
The reason actual, human translators are so useful is because they master languages in order to make sure subtleties are not lost. Words develop entirely new meanings depending on regions and social status. Speaking from personal experience, a person from New England using the word "wicked" in a sentence is not using a dictionary definition of wicked (unless, in fact, they are using the dictionary definition of wicked). A translator can recognize dialects and slang, guaranteeing that there are no misunderstandings. Any device that wants to be nearly as effective as a human translator needs to be able to understand the context of each word depending on the region and its position in the sentence.
A device that would translate speech would also need to be able to deal with incredibly thick accents. Even native speakers will have their own way of using their language. There are clear differences in how someone from Boston speaks when compared to someone from the South, or even between different cities in the same state. In countries that don't have such widespread communication, the ones were translators are needed most, accents can sound like another language even if they aren't using a different dialect.
Slang is an entirely different beast for a device that performs translation to deal with. They can be entire phrases that aren't supposed to be taken literally (a lounge lizard was not a reptile) or words that are used to mean something other than the definition (the wicked example). Each dialect can have its own slang, and being able to distinguish between dialects and whether or not a word is being used as slang are skills a human translator would have that a machine would have difficultly replicating.
In a place where any slight error in communication can lead to a loss of life, it's important that we don't forget just how complicated language is. There's a reason human translators are still an important part of diplomatic relations and businesses.
Raytheon BBN has been given the almost-impossible task of developing a device that can perform two-way speech-to-speech translations, among other things. Now, I don't doubt a comprehensive translator can come from this, I doubt that any device can effectively translate human communication.
The reason actual, human translators are so useful is because they master languages in order to make sure subtleties are not lost. Words develop entirely new meanings depending on regions and social status. Speaking from personal experience, a person from New England using the word "wicked" in a sentence is not using a dictionary definition of wicked (unless, in fact, they are using the dictionary definition of wicked). A translator can recognize dialects and slang, guaranteeing that there are no misunderstandings. Any device that wants to be nearly as effective as a human translator needs to be able to understand the context of each word depending on the region and its position in the sentence.
A device that would translate speech would also need to be able to deal with incredibly thick accents. Even native speakers will have their own way of using their language. There are clear differences in how someone from Boston speaks when compared to someone from the South, or even between different cities in the same state. In countries that don't have such widespread communication, the ones were translators are needed most, accents can sound like another language even if they aren't using a different dialect.
Slang is an entirely different beast for a device that performs translation to deal with. They can be entire phrases that aren't supposed to be taken literally (a lounge lizard was not a reptile) or words that are used to mean something other than the definition (the wicked example). Each dialect can have its own slang, and being able to distinguish between dialects and whether or not a word is being used as slang are skills a human translator would have that a machine would have difficultly replicating.
In a place where any slight error in communication can lead to a loss of life, it's important that we don't forget just how complicated language is. There's a reason human translators are still an important part of diplomatic relations and businesses.
John Keller is editor-in-chief of Military & Aerospace Electronics magazine, which provides extensive coverage and analysis of enabling electronic and optoelectronic technologies in military, space, and commercial aviation applications. A member of the Military & Aerospace Electronics staff since the magazine's founding in 1989, Mr. Keller took over as chief editor in 1995.
