In this paper, we use the data of China’s manufacturing listed companies from 2009 to 2018, adopt the method of propensity score matching and double difference (PSM-DID) to solve the sample’s selective bias, and select the accelerated depreciation policy of fixed assets issued by China in 2014 as a quasi-natural experiment to verify the robustness of the empirical results, which will affect the R&D investment of manufacturing enterprises and the structural tax reduction of China. This paper makes an empirical study on the effect of fixed asset investment to restrain the financialization of manufacturing enterprises. The results show that (1) accelerated depreciation policy of fixed assets significantly promotes the R&D investment and fixed asset investment of enterprises and reduces the level of enterprise financialization; (2) accelerated depreciation of fixed asset local tax policy, through guiding the R&D investment, fixed asset investment, and deferred income tax acquisition of enterprises. It guides the investment of enterprises to the real economic field, thus reducing the financial assets of enterprises. The investment has restrained the financial trend of real enterprises. The conclusion of this paper is of practical significance to support the formulation and implementation of the national structural tax reduction policy and to clarify the regulatory role and mechanism of the structural tax reduction policy.

The multi-mode matching has noteworthy transformations equated with the classical multi-mode matching algorithms. It is frequently used for the policy part of the TCP connection to connect the English characters. In this article, we analyzed the features of multi-mode similarity for audit information retrieval in a cluttered environment. The proposed model analyzed the performance theorem of a multi-mode matching algorithm for audit information retrieval. It also analyzed the shortcomings of existing multi-mode similarity systems and proposed a multi-mode algorithm based on the trail hash trie matching machine suitable for mixed Chinese and English environments. The algorithm converts the set of pattern strings into multiple finite automata and then builds a state driver using the set of pattern strings. The state driver is driven by the characters of the string to be matched in turn, and each finite automaton is driven by the state driver to achieve similar multimodal matching with mixed English and Chinese characters by allowing the insertion errors. The algorithm does not need to match every character and can make full use of the information of this unsuccessful match during the matching process and skip as many characters as possible by combining the improved text window mechanism. It can control the upper limit of allowed errors for each pattern string. The matching speed is independent of the number k of allowed insertion errors. The algorithm has comprehensive application projections in the fields of information auditing, database, and information retrieval, respectively.