Corpus data have emerged as the raw data/benchmark for several NLP applications. Corpus is described as a large body of linguistic evidence composed of attested language use. It may be contrasted against sentences constructed from metalinguist reflection upon language use, rather than as a result of communication in context. Corpus can be both spoken and written. It can be categorized as follows: monolingual, representing one language; comparable, using multiple monolingual corpora to create a comparative framework; parallel corpora, wherein, corpus of one language is considered, and the data obtained, is translated in other languages. The choice of corpus depends on the research question/the chosen application. Adding linguistic information can enhance a corpus. Analysts, human or mechanical, or a combination achieves annotation. The modern computerized corpus has been in vogue only since the 1940s. Ever since, the volume of corpus banks have risen steadily and assumed an increasingly multilingual nature.