Abstract: In the process of agricultural vertical search engine research, Chinese word segmentation is an important research direction. Vertical search engines existed inaccuracy, slow velocity and other shortcomings for information extraction based on traditional agricultural. In this paper, the Trie tree method was adopted as the basic model to design the word segmentation dictionary specifically for agricultural use based on MySQL database. The word segmentation dictionary could make full use of the database for dictionary. It could be a thesaurus remote sharing and common maintenance, convenient access to different system. In the dictionary, used the term in Chinese location code and database table row number corresponding to the acronym, and according to the needs of agricultural vertical search engine, set up agricultural word part of speech coding. This dictionary stored the double array Trie tree according to the classification of storage structure. It could reduce the memory space of construction effectively. At the same time, the agricultural word segmentation dictionary structure also had reference significance to other field.