Multi-view clustering-based multilingual data pattern mining has received significant attention in recent years due to its ability to fully leverage the complementary and consistent information from multiple languages.Although existing methods achieve educational toys encouraging performance, they often jointly optimize representation learning and pattern mining within a single feature space, which may degrade the effectiveness of multilingual data pattern mining.To address this issue, this paper proposes a multi-granularity contrastive learning-based deep multilingual data pattern mining method here (MCL), which consists of three view-invariant learning modules: structure learning, semantics learning, and partitioning learning.
MCL integrates these three levels of view-invariant learning into an end-to-end framework, comprehensively exploiting the consistency and complementarity of multi-view data, thereby significantly improving the accuracy and robustness of multilingual data pattern mining.Finally, through extensive experiments on five datasets, MCL shows to establish a new benchmark for ACC, NMI, and PUR, proving its superiority and effectiveness.