Translating is an activity involving two languages. Language has conventions to allow people to communicate with each other smoothly. A person grows up in a certain discourse community learning the language through exposure. However, it is difficult for human minds to recall the norms of language in a conscious way. Fortunately, now that the computer network is prevalent, a lot of language data can be analyzed by a computer program to extract frequencies and usages and present rules applicable to a particular discourse community. For a professional translator, the Internet has become an increasingly indispensable resource. This article first discusses the norms of language and explores the issues of standard usages for a certain discourse community. Examples are used to demonstrate how word usages and collocations can be extracted from the Internet to facilitate the recognition of preferred linguistic conventions. The article then goes on to consider the concept of "norm" in translation studies, emphasising the role Internet plays in the process of translating. More examples are used from results of web search to show how observations of authentic usages can facilitate the choices of words, collocations, grammatical structures, and phraseological units to conform to the conventions of the target language. Finally, this article proposes a model of translation based on the Web as Corpus, explaining what roles language users can play in the formulation of language conventions, as well as what the translator's roles and responsibilities are in this machinery. [ABSTRACT FROM AUTHOR]