In information retrieval systems, it is very important that indexing is defined very well by appropriate terms about documents. In this paper, we propose a simple retrieval model based on terms distribution characteristics besides term frequency in documents. We define the keywords distribution characteristics using a statistics, standard deviation. We can extract document keywords that term frequency is great and standard deviation is great. And if term frequency is great and standard deviation is small, the terms can be defined as paragraph keywords. Applying our proposed retrieval model we can search many documents or knowledge using the document keywords and paragraph keywords.