[Objective]In order to further improve the depth of the semantic relevance mining of policy text, this paper explores the method of policy text relevance mining.

[Methods] This research integrates dependency parsing analysis and word embedding model to mine the deep semantic relevance of policy text content from the perspective of sentence information and word meaning information, and it fully considers the language characteristics of the policy text when setting the dependency syntax extraction rules.

[Results]In terms of algorithm effect, in the test data set with a relatively low degree of policy text association, the algorithm F1 value reached 0.857, which is an increase of 22.78% compared to the traditional conventional algorithm; in terms of algorithm function, the policy text relevance can be described from the subtle differences in words.

[Limitations]In semantic information mining, the algorithm currently uses an open source model, which can subsequently independently train word vector models in specific policy domains to further improve accuracy; In sentence information mining, the algorithm relies on the accuracy of existing dependency syntactic analysis tools.

[Conclusions]The algorithm proposed in this paper has good effects and strong functions. It can effectively reveal the degree of policy text association. So it can bring new research perspectives and tools for quantitative research on policy text.