基于核方法的敏感信息过滤的研究

Sensitive information filtering based on kernel method

在深入分析敏感信息过滤任务的特点和难点的基础上,针对现有一般的信息过滤方法的不足,提出了一种利用敏感词的组合信息来改进过滤效果的思想。进而,研究了在核方法的框架下特征共现行为建模的原则并提出了复合ANOVA核来刻画特征组合行为。通过真实信息过滤环境中的测试评估,显示了此敏感信息过滤方法的有效性。

Based on the discussion about the characteristics and the hardness of sensitive information filtering task, the idea of employing the combination of sensitive words to refine the filtering effect was proposed. The compound-ANOVA kernel to modeling the combination of features was introduced under the kernel method framework. According to the evaluation in the real settings, the method exhibits its superiority on the sensitive information filtering task.