Mining Semantically Consistent Patterns for Cross view data with CCA and CJFL

Abstract

We often faces the situation that the same semantic concept can be expressed using different views with similar information, in some real world applications such as Information Retrieval and Data classification. So it becomes necessary for those applications to obtain a certain Semantically Consistent Patterns (SCP) for cross-view data, which embeds the complementary information from different views. However, eliminating heterogeneity among cross-view representationsis a significant challenge in mining the SCP. The existing work has proposed the effective Isomorphic Relevant Redundant Transformation (IRRT) and Correlation-based Joint Feature Learning (CJFL) method for mining SCP from cross-view data representation. Even though existing system uses the IRRT for SCP from low level to mid-level feature extraction. Some redundant data and noise remains in it. To remove redundant information and noise from mid- level feature space to high level feature space, CJFL algorithm is used. We are using Canonical correlation analysis (CCA) method instead of complex IRRT which also lags to remove the noise and redundant information.