In everyday life, computer is not only supposed to understand computing language which programmer or other professional input into computer. It also need to understand the language we use in our daily life so that people who are not a professional can easily use the computer and the computer can understand what to do the next. Natural Language Processing is part of the Auto Intelligence study.

 

The problems with creating a AI system that understand the language and knowledge from the outside world is that they need enormous amount of experience to make sure that understanding is complete so that AI can distinguish what is supposed to be right and what is supposed to be wrong.

 

Here are some functions we use that involving Natural Language Processing:

 

Text to Speech and Speech synthesis; speech recognition; Parsing; information retrieval; text-proofing; machine translation; question answering.

 

The translation of one language to another involve huge amount of works. It require the AI to understand the meaning from one inputting language and translate into another language with accurate translation. In the class, we learned from the representation that it is not always accurate from what we translate and what we get from the computer AI. The reason behind is that computer do not accumulate experience. (When I say experience, I mean they do not build a concept over and over).

 

This is not the only problem with Natural Language Processing. When processing language, more possibilities will result in bad result. For example, words in Chinese usually involving multiple meanings when used in different context. It is very hard to write a program that tells which one to translate at what time.

 

Secondly, accent and habit of language using is also a hard problem, people within specific region develop their own ways of saying things, but such accent or habit is unpredicted by computer due to the fact that computer AI is only able to understand standard language.

 

Finally, the way to respond is too straight forward and sometimes unexpected result will be given when using natural language processing. For example, when we input an question like how many student passed the course last year? We may get an answer that says 0, but the truth maybe that this course is not opened last year.

 

To sum up, Natural Language Processing still has a long way to go, because it does involving machine leaning and it requires computer to process huge amount of examples to accumulate experience so that mistakes can be lessen