Processing of natural language is branch of linguistics, artificial intelligence & computer science and its purpose is to have interaction among natural language of human beings and computers. We can say it is related to field of computer–human interaction. There are different challenges in this field like understanding of natural language i.e. allowing machines to have understanding from natural language of human beings. Mostly available tasks of natural language processing are: analysis of discourse, morphological separation, machine translation, generation and understanding of natural language, recognition of named entities, part of speech tagging, recognition of optical characters, recognition of speech and analysis of sentiments etc. Current research in NLP is showing more interest on learning algorithms which are either unsupervised or semi-supervised in nature. These techniques of learning can perform this task of learning from data which is not annotated manually with required answers or by applying mixture of non-annotated & annotated data. Normally, this job is very hard as compared to learning which is supervised & usually shows little correct results for particular amount of data as input. But there is large quantity of data is available which is non annotated in nature i.e. whole contents available on world wide web and it normally produces less accurate results. This paper discusses about a survey of different techniques of natural language processing.