As text mining techniques evolve, keyword extraction techniques are critical to text mining research that handles large amounts of documents such as document searches, clustering, classification, indexes, and etc.. However, in an existing keyword extraction study, a pre-treatment process is required to extract the critical compound noun as a single keyword. And in the case of Web documents, since the era of Web 2.0 is coming, many users create their own documents. This is a difficulty in constructing the word dictionary for each document due to the problem of the amount of documents and the problem that the field can not be specified. So existing studies have led to research to extract keyword extraction without a pre-treatment process for the compound noun. In this paper, a study is conducted to find a compound noun that can be treated as a single keyword within a document without pre-treatment to each document. For this purpose, use the PMI(Point-wise Mutual Information) value to find the connection between the words in the document. The results of this study are expected to lead to better outcomes for research, such as document retrieval, document summaries and topical detection.