A compatible model of unstructured data for cross-media retrieval in the field of tourism

Abstract

With the development of multimedia, the amount of unstructured data of multimedia is increasing, especially for the tourism. Meanwhile, it is one of the key issue to analysis large-scale unstructured data, which helps us to find the hidden relevance between redundant and different data. How to retrieve efficiently, and recommend accurately for cross-media retrieval is more and more important. This paper proposes a new data mode for cross-media retrieval - unstructured data compatible model, short for UDC model. The UDC model is constructed by its own metadata. All metadata are organized by a certain hierarchical relationship. Every metadata consists of three layers: the feature layer, the semantic layer and the compatibility layer. Furthermore, this paper presents retrieval and recommendation algorithms based on UDC model. The experiment results demonstrate that the retrieval engine based on UDC mode can be more effective for cross-media retrieval and recommendation.