With the need of international dissemination of Chinese culture, the problem of translating traditional Chinese texts gradually emerges. The study embeds a computer semantic model into the English translation of The Analects of Confucius, and constructs a natural language understanding model based on S-LSTM network through semantic representation of natural language processing. In order to explore the performance of the S-LSTM model, it is compared with RNN, LSTM, I-LSTM and other models in terms of training time and accuracy, so as to validate the superiority of the S-LSTM model in this paper. This paper deeply explores the philosophical connotation of the character “body” in The Analects, and studies the structural complexity of the translation of the character “body” through the S-LSTM model. Finally, the English translation strategy of The Analects and other classics is proposed. Among all the comparison models, the S-LSTM model has the fastest training speed and the highest accuracy. The translation of the word “body” in The Analects and the local complexity of the ministry are characterized by complication. The local complexity of the noun and the subject in the source English language, and the overall complexity of the “be-passive” structure have obvious effects on the structure of the translated Chinese character “body”.

Along with the development of the times, online classroom teaching activities have been carried out in different degrees and frequencies in various schools, and the gradual advancement of education informatization has improved the software and hardware environment of online classroom and other forms of teaching. The study designed a 21-item questionnaire related to English online classroom learning and selected all the students who participated in English online classroom teaching in a school for the survey. After collecting the questionnaire data, factor analysis and multiple stepwise regression model were used to conduct multivariate statistical analysis on the English online classroom data. And on this basis, the teaching plan was adjusted according to the actual learning behaviors of the high, medium and low risk level students themselves respectively to achieve personalized teaching. The results show that students’ satisfaction with the English online classroom is high, and that pre-course homework analysis, group learning, formative learning evaluation, students’ independent learning ability and online learning resources are the key positive factors affecting the learning effect of the English online classroom, with the influence coefficients of 0.036, 0.055, 0.048, 0.044, and 0.062, respectively. At the same time, after the optimization of teaching strategies, the students’ logged-in learning behavior, participation rate in interactive test questions and grades were significantly improved, proving the effectiveness of the strategy.

In today’s era, the transformative power of computing is highlighted, and computational thinking has become the core literacy and essential ability of learners, while computer education is an effective carrier for cultivating computational thinking. The article firstly researches the theory related to collaborative filtering and generative adversarial recommender system. Then it combines SeqGAN with traditional CF algorithms, proposes to use sequence generative adversarial network for missing data prediction, and makes appropriate improvements to SeqGAN to make it suitable for generating scoring data, and then further designs a computer teaching system based on this model. The article launches performance testing experiments on Ali’s real dataset UserBehavior, and conducts experiments on the effect of computer education with the students of computer application major in a secondary school as the research object. The results of the study show that in the comparative analysis of the pre-test and post-test of computational thinking of the experimental class, the mean of the total score of computational thinking of the experimental class in the pre-test and post-test is 71.17 and 78.35, respectively, and the post-test is more than 7 points higher than the pre-test. It can be concluded that the teaching model of multilevel computational modeling designed in this paper promotes the development of students’ computational thinking and academic performance, improves students’ learning attitudes, and increases classroom participation.

With economic globalization and the increasing complexity of inter-enterprise business linkages, corporate financial systems have gradually taken on the characteristics of complex networks. This paper firstly gives an overview of the complex network and introduces its basic topological properties, such as clustering coefficient and path length. After that, through the principal component analysis method, the enterprise financial risk early warning indicators are identified, and the key indicators are screened to improve the early warning accuracy. Based on these properties, the financial risk conduction network model of complex enterprises is constructed, the characteristics of the network are analyzed, including network density, centrality distribution, etc., and the effect of financial efficiency enhancement of complex enterprises under the optimization of topology computation is verified in real cases. The results show that most of the financial risk indicators of enterprises have strong correlation, and the degree of centrality of 9 indicators such as “gearing ratio and quick ratio” is more than 50%. In addition, the indicators of “current asset turnover ratio, interest coverage multiple, net profit growth rate” can play the role of intermediary and bridge, and the risk transmission effect among the indicators is high. The threshold value of 0.65 is the watershed of the changes in the financial structure of enterprises, and most of the financial risks in the network have a high degree of similarity in the financial structure when the degree value is 70, and it is negatively correlated with the coefficient of agglomeration, and the coefficient of agglomeration decreases with the increase in the intensity of the points.

In order to explore the deficiencies in the teaching process of marketing majors in higher vocational colleges and further improve the teaching quality of marketing majors in higher vocational colleges. This paper utilizes the improved ID3 algorithm to construct the SLIQ data mining algorithm to improve the teaching quality of teachers of marketing majors in higher vocational colleges and universities. Using ID3 algorithm to build a decision tree to get the portraits of teachers and students, at the same time, in order to reduce the computational complexity of ID3 algorithm and the problem of multi-value bias, the concept of sample structure vector similarity is introduced, and the degree of information gain is optimized to get a more reasonable decision tree. On this basis, based on the improved ID3 data mining algorithm, a teaching quality assessment system for senior marketing majors based on SLIQ algorithm is designed, which identifies important factors affecting teachers’ teaching quality by mining a large amount of data in the teaching process.The AUC value of the SLIQ data mining algorithm is 0.98, which can effectively improve the algorithm’s generalization ability, and it has an excellent performance in the teaching quality assessment task. The performance is excellent. In this paper, we systematically identify “the principles of marketing” and “the degree of seriousness of teachers’ homework correction” as the key factors to improve the teaching quality of marketing teachers. It provides a scientific basis for improving the quality of teachers’ teaching.