Abstract: [Purpose / Significance] In order to deeply utilize fishery patent literature in "fishery knowledge service system", we construct a specialized taxonomy for the information organization and test it as an indexing tool in the system. [Method / Process] First, 10, 323 metadata of Chinese fishery patents are selected and analyzed in terms of 4-digit and 6-digit IPC numbers frequency. Based on the results, 12 IPC classification codes are chosen as categories in fishery patent literature classification. Then, based on the analysis of patent titles, it is proposed that a patent title consists of three kinds of phrases: those of industry attributes, of business types, and of invention types, among which phrases of business types are the most applicable for classifying and indexing patent literature. In addition, the subject terms and the ending parts of subject terms are analyzed and the high-frequency subject terms and words of each business type are ranked. Last, the strategy of indexing fishery patent literature by using IPC classification numbers and high frequency subject terms is put forward, and as a result, most of the fishery patent literature is classified properly with the assistance of computers. [Results / Conclusions] Giving that the total accuracy rate of machine indexing is 91.44%, and the missing rate is 7.94%, with the comparison between the manually classified and machine classified sampling data of patent literature in 2016, it is concluded that the goal of this study is achieved. The research shows that the classification system constructed from fishery patent literature is practical and the indexing strategy has a high application value.

Key words: fishery, patent literature, classification, machine indexing