字幕列表 影片播放
Over the years, since computer science evolutions, Artificial Intelligence principles demanded
the Natural Language Processing so that machines or software could communicate efficiently
with human. Today we will make an introduction to Natural
Language Processing and furthermore we will present chat bots, which are directly connected
to Natural Language Processing.
So what is Artificial Intelligence exactly? John McCarthy in 1956 stated that Artificial
Intelligence is “the science and engineering of making intelligence machines”. In simple
words, Artificial Intelligence is the Computer Science sector that has to do with the design
and implementation of software that are capable to imitate the human cognitive skills, showing
characteristics that are normally attributed to human behavior. Some examples are the problems
solution, vision, learning, conclusions and understanding the natural language.
Natural Language Processing is a field of computer science and more specifically is
a field of artificial intelligence but also is included in linguistics concerned with
the interactions between computers and human languages.
The main purpose is to create and implement computational models, that are able to extract
meaningful information from natural language input and furthermore to produce a natural
language output. Some aspects of Natural language processing have to do with communicating with
the computer, the machine translation and the browsing and filtering of texts, written
in natural language from an agent. Natural Language Processing is actually identical
to the field of computational linguistics which is divided to theoretical and applied
part. The theoretical computational linguistic, deals with how people acquire and use knowledge,
to produce and understand natural language. The applied part focuses on the practical
results of modeling the use, of human language. After processing the natural language we can
understand more about the world. And if scientists manage to succeed in the creation of a computational
language model, we will own an extremely strong communication tool.
Natural language is divided to the written and oral language. In order to conquer oral
language, it is essential to understand how written word is used and structured.
Alan Turing is a famous scientist that is known as the “father” of computer science
and artificial intelligence. In 1950 he published his famous article titled: Computing Machinery
and Intelligence which introduced a criterion of machine intelligence, what today is called
Turing Test. The idea was to test if a human could not tell apart if he/she was communicating
with a machine or a human. Practical work began with the Georgetown-IBM
experiment in 1954, which was developed by the Georgetown University and IBM. This experiment
involved fully automatic translation of more than sixty Russian sentences into English.
Although at first, scientists were very optimistic about the results of the experiment and thought
that the machine translation would be a solved problem within three of five years, real progress
was actually a lot slower than expected. Moving forward in 1972, T. Winograd created
the system SHRDLU, using the language Lisp. This is a natural language system working
in restricted “blocks worlds” with restricted vocabularies. Examples of its aspects are
interpretation of questions, states and directions, also ability of entailment and learning new
words. During the 70’s many programmers began to
write “conceptual ontologies” which structured real world information into computer understandable
data. Up to the 80’s most of Natural Language Processing systems used hand written rules.
In the late 80’s though, of machine learning algorithms were introduce so the systems could
use decision trees or statistical models which make soft, probabilistic decisions based on
attaching real-valued weights to the input data. Nowadays systems focus on unsupervised
or semi-supervised learning algorithms that are able to learn from data that has not been
hand-annotated. In order to achieve natural language processing
it is useful to divide the process to some steps and parts.
First, a morphological analysis has to be done, where separate words are analyses to
their components and the non verbal symbols. Then, we proceed to syntactic analysis where
linear sequences of words are converted into structures that illustrate how words are connected
to each other. Some sequences may be rejected by the system, if they violate some of the
language rules. Semantic Analysis follows, which gives meaning
to the structures that are produced from the syntactic analysis.
Then we must do a discourse Integration, which analyzes the meaning of a single sentence
compared with the meaning of the previous and the next sentences.
And at last we apply a pragmatic analysis in order to define the real meaning of the
structure that represents what it has been said.
During Natural Language Processing several problems may have to be encountered. These
problems have to do with the fact that people often use natural language in a certain way
expressing feelings or questions without using the official language and without satisfy
every rule of the language. For example, the sentences of natural language
contain incomplete description of the information that intends to transfer. We may say, “I
called Maria to go to the movies and she agreed” but we actually mean, “Maria was at home
when I called her. She answered the phone and I asked her to go to the movies. She said
ok.” Another example is that the same sentence
may be used for different occasions and meanings and on the other hand there are a lot of ways
to say the same thing. For example we may say “Maria’s birthday is on the 1st of
September” or “Maria was born on the 1st of September”.
And also there is the fact that none program of natural language can be ever complete since
new words, new phrases or meanings are often created by people.
Natural Language processing in our times is widely spread and used. A lot of application,
software and machines all over the world use this kind of processing in order to achieve
high quality communication. There is machine translation that automatically
translates text from one human language to another.
Also Natural language generation and understanding systems exist, that have to convert information
from computer databases into readable human language and understand several input from
users. Optical character recognition has to determine
a corresponding text given an image representing printed text.
Question answering has to do with answering human language questions. There are specific
answers to some questions if we ask for example the capital of a country. But some questions
can have different answers for example “what is the meaning of life”.
Speech recognition where given a sound clip of a person speaking, the system has to determine
the textual representation of the speech. As you can see, natural language processing
is really important for today’s applications and further evolution of technology. It helps
simple and more complex systems to achieve their goals.