With the development of economic globalization, international tax problems such as international double taxation and international tax evasion continue to emerge, and global tax governance is receiving increasing attention and focus from the international community. This paper, based on the standard framework of information exchange in international tax law, provides a preliminary analysis of the effectiveness and function of international tax law. The linear transformation of the data input sequence leads to the acquisition of corresponding vector values, and the international tax law topic faceted tree is constructed based on the knowledge extraction results. Using the BERT model, the semantic information text matrix for international tax law is obtained. Extract and mark the relevant elements of the international tax law information exchange standard, combined with element feature marking. The extraction results are analyzed in terms of evolution and outcome: from 2008 to 2019, China signed the Convention and tax information exchange agreements with 33 and 10 countries, respectively, accounting for no more than 50% of the total. Between 0.0178 and 0.0187 is the size of the coefficient of the policy effect of information exchange agreements on cross-border tax avoidance of businesses. This means that it can stop businesses from avoiding cross-border taxes after the agreements go into effect, but it has a pretty small effect.