#1.1 intro NLP learning
#This chapter is first class in NLP course of the University of Sydney. It introduces some interesting challenges of NLP from ethical respective. In the future, I might give more in algorithm and programing.
#The blog is trying to record the idea from the class and practicing writing skill which might not accurate enough for a research or coding. But it still useful because it might provide some insteresting ideas and some sumarries of knowlege to you guys.
Nature Language Process is a process which transfers nature language to machines especially to computers. This is a simple definition of NLP, the real meanings are more than it. NLP is trying to make machines understand the demands from humans by following human’s instructions. These instructions could be speaking, typing and movements. However, most of computers are made by electronic tubes which is totally different from our human’s brain. In this case, neural networks are put forward in machine learning to let computer imitate thinking process of real human. These networks do well in many cases. But there are still some challenges for NLP.
Ambiguity of humans language is considered as one of biggest disadvantages to NLP. This is because computer are accurate when they processing data. But in the real life, daily conversations between humans are ambiguity which could not provide accurate information to the computer. Thus, when the computer receive the distort information it could not give right feedbacks even through the algorithm behind is advanced. This challenge occurs in many cases and almost all languages in this world. However, this is hard to be solved because the language human speaking can not be restricted by computer rule. That means the ambiguity among real conversations could not be avoided, which has tremendously adversed impacts on the NLP. For example, when human say “I miss you”. There are two meanings, “I miss you” could be “I love you and I want you come back”. It also could be “I just miss you, but it does not mean we can get back”. So, for the computer’s processing, if computer can not get the real meaning of
“I miss you” it will give the totally wrong response to speaker. In this case, one possible solution is to get into the context of “I miss you”. However, NLP is different from mechine learning due to the limitation of abundant data from real conversation between humans to training its neural networks.
In addition to the ambiguity of humans language, the respect between two speakers is also challenging the NLP with artificial intelligence. In real conversation, human show their respects to each other which could make conversation successful. But when human talk to robots, speakers know the object they talking with is not real so the language they use is pretty bad and the content they talk about might be strange. For example, in real world we don’t ask people questions directly because this is not polite. But when speaking to computer speaker can speaking everything what they want to say. In this case, for human they might get angry when they are treated in an inpolite way. But for computers, they won’t get angry at all, even through they can nobody cares. This is an obstacle for computer to process the humans language because in many cases the conversation between two speakers are unreasonable.