| Text-based
Word
Processing
Most users of word processors and similar programs don’t think
twice about the wizards, grammar checks, and trouble-shooting menus.
However, NLP is present in something as common as Microsoft Word.
Some of the natural language processing done in word processing,
such as spell checking and suggestions, is so simple that it isn’t
even considered to be NLP any longer. Other programs, such as the
grammar check, are still in a state of early development.
Information
extraction and generation/retrieval
Froogle
While searching for products previously meant looking through a
database of items known to be for sale, Google brought online shopping
to a new level by incorporating natural language processing. Froogle
scours the web for goods, and it uses natural language processing
to determine which items are being sold on a website. How does a
computer figure out if a certain company is offering a product?
How can it parse the data for the price, description, and picture?
NLP’s statistical and linguistic tools help us with these
problems. Although more complicated than most online shopping companies,
Froogle has the power of the entire World Wide Web behind it.
Intellishrink
Microsoft makes a plug-in for Microsoft Outlook that ships with
Microsoft Outlook Mobile Messenger. The program identifies important
e-mail messages and modifies them to fit on the smaller screen of
a PDA. This application of NLP allows a computer to interpret how
important your messages are, using intelligent “sense”
to process your incoming data.
Chatterbots
Chatterbots have had their fame among the general population. These
programs can answer questions and engage in simple dialogues with
the users. They parse the statements or questions and respond to
simulate conversation. Most AOL Instant Messenger (AIM) users have
had at least a conversation or two with “SmarterChild,”
the one buddy who is always online. Users can ask SmarterChild about
itself, and can ask personal questions as well as query about items
of general interest, such as stocks and current events. Alice
is a more widely known chatterbot. http://www.alicebot.org/ A sample
conversation might look like this:
Human: Hi
Alice
ALICE: Hello there.
Human: How are you?
ALICE: Everything is running smoothly.
Human: That's good
ALICE: How are you doing?
Human: Great
These
chatterbots are some of the most common and apparent types of AI
the general population is aware of. Natural language processing
allows people to talk to the machines using not C or Java, but English,
French, or Chinese.
Forecast
Generator (FoG)
There are many translation applications that involve natural language
processing, but NLP can be applied so that translation isn’t
even necessary. The Environment Canada, to help weather forecasters
increase productivity, developed FoG. Instead of modifying a map
showing weather data and then translating the textual report from
English to French, FoG can directly generate weather reports in
both English and French, eliminating downtime.
Translation
Microsoft, IBM, etc
Almost every student who’s taken a language in high school
remembers using AltaVista’s Babelfish
to translate some portion of her French or Spanish homework assignment.
Translation through computers is fast, easy, and cheap. However,
although it might seem simpleenough, companies throw a lot of money
each year at this project to improve accuracy. Users of Babelfish
know that they cannot type their entire essays in English and expect
the French or Spanish version to translate them coherently because
of accuracy issues.
Microsoft, for example, focuses on machine translation as one of
its major projects in NLP. Their technology utilizes existing translated
text to learn translation.
CMU
Babylon
Carnegie Mellon University’s Babylon does speech-to-speech
translation from Arabic to English and from English to Arabic on
PDAs. It may be some time before technology can advance to the state
of the devices the Martian used on those Saturday morning cartoons,
but research is advancing.
|
