"...NLP is the use of computers to understand human "natural languages"such as
English French & Japanese. By understand we do not mean that the computer has
human like thoughts feeling and knowledge. We mean only that the computer can
recognize and use information expressed in human languages."
--Michael Covington
"...NLP is the ability of a computer to process the same language spoken or written
that humans use in normal discourse on the other hand is that field of study within
cognitive science which deals with linguistic behavior."
--Klaus K. Obermeier
Figure 1. shows the classic six areas natural language processing (nlp) research.
The field of natural language processing is often divided into six areas.
The generation of speech by computer.
"...the creation of speech by computer, is considerably easier than speech recognition.
A synthesizer's most important task is to simulate coarticulation. This is usually done
by providing several allophones for each phoneme, each to be used when a different kind
of phoneme is adjacent. Recognizable speech has been generated by this method since 1960's."
--"Natural Language Processing for Prolog Programmers" by Michael A. Covington
"Phonology is the analysis and recognition of speech sounds. The phonology component of an
NLP system translates a continuum of sound waves into a string of discrete symbols chosen
from a seAt of sounds called phonemes used by a particular language."
--"Prolog Programming in Depth" by Michael A. Covington, Donald Nute and Andre Vellino
A program that allows you to communicate with a computer using a natural language. An NLI typically
includes both the natural language understanding and speech generation capabilities. These programs
are known as natural language front-ends.
"...The task for speech recognition systems is to separate the signals and classify them
into groups of sounds that make up individual words. ...transforms the analog acoustic
data into discrete digital format."
...Speech recognition systems are based on three different approaches:
- a template matching approach in which speech is reduced to a number of stock
tokens that need to be identified.
- an acoustic feature identification approach
- a multi-level approach in which syntactic and semantic information drives
the analysis.
--"Natural Language Processing Technologies in Artificial Intelligence"
--Klaus K. Obermeier
An area of AI research that is attempting to use computers to translate text from one language
to another. These programs often use a combination of natural language understanding and speech
generation.
"...performs the transformation of linguistic input from a source language into a
target language."
"...the direct approach consists of mapping from source to target language without
intermediate representation."
--"Natural Language Processing Technologies in Artificial Intelligence"
--Klaus K. Obermeier
The part of NLP research which is investigating methods of allowing computers to understand instructions
given in a natural language.We measure the level of understanding on results. For example, if we asked
someone to perform some action, if the person performs the action, we feel we were understood. If the person
performs an action incorrectly then we feel that the person did do fully understand us. If the person
does not perform any action then we feel that we were not understood at all. A goal of natural language
understanding is investigating methods to allow the computer to understand enough of a command given to
it in a natural language to perform some action.
There are a number of problems that must be addressed in attempting to get an computer to understand
natural language:
- Ambiguity
- multiple word meanings
- syntactic ambiguity
- unclear antecedents
- Imprecision
- Incompleteness
- Inaccuracy
Ambiguity can be overcome by putting an idea in context. Imprecision can be overcome by relating the
idea to a familiar situation. Incompleteness can be overcome by completing the idea based on our
expectations of likely events. Inaccuracy can be overcome by inferring the intended meaning by recognizing
a familiar pattern.
Text generation determines what a computer says and how the computer says it.
"...text generation has two phases: during the first phase, the content and structure of the
discourse is determined by means of a strategic component; during the second phase,
the message is transposed into NL via tactical component. ...Text generation programs
construct the text based on a communicative goal. The strategic level is the level where all
the planning takes place, i.e., the selection of concepts and their relationship. On the
strategic level, the problem solving of the program includes also the planning of utterances
with respect to what the informative content of the utterance is and the level of communication
with the user of a particular system. On the tactical level, the program has to fine-tune
the utterance depending on content, length, and stylistic variations.
--"Natural Language Processing Technologies in Artificial Intelligence"
--Klaus K. Obermeier
| 
