D1. 4.1 b Kernel Canonical Correlation Analysis

Kernel Canonical Correlation Analysis (KCCA) is a technique for finding common semantic features between different views of data. It can be used for finding semantics from different languages that share the same meaning. This information can then be used for mining databases with multilingual text documents. In this report we present theory behind KCCA and application of it on cross-language information retrieval and classification