Term proximity has been successfully used in many entity retrieval searches and enhance the quality of the retrieval systems. In general, the goal of entity searches is to retrieve a ranked list of entities in response to a user’s query. The entities could be organisations, products, location, or people. Some of the proximity models that were successful used association discovery in a window of text rather than in a whole document. All current studies have only investigated fixed window sizes; as such, we propose an adaptive window size approach for proximity searches. In this study, we concentrated on a particular type of entity search: expert finding. We used some of the document’s attributes such as document length, average sentence length, and number of candidates in the document to adjust the window size of the document. The results of the experiments indicated that considering the document’s features when determining the window size did have an effect on the effectiveness of retrieval and provided much better results than a range of baseline approaches using fixed window sizes.