Global Multi-modal Knowledge Research Center develops the linguistic resources and techniques for opinion analysis. Based on preliminary studies on multilingual linguistic properties, we develop semi-automatic methodologies to construct annotated corpora amd sentiment dictionaries. As engineering research, we have built up a testbed for crawing and cleaning a huge amount of multilingual data on the world-wide web. Also, we have established a learning system based on a shallow machine learning technique. In addition, we develop parallel and high-performance storage techniques for large-scale data storing and retrieval.