Based on the text orientation classification, a new measurement approach to semantic orientation of words was proposed. According to the integrated and detailed definition of words in HowNet, seed sets including the w...Based on the text orientation classification, a new measurement approach to semantic orientation of words was proposed. According to the integrated and detailed definition of words in HowNet, seed sets including the words with intense orientations were built up. The orientation similarity between the seed words and the given word was then calculated using the sentiment weight priority to recognize the semantic orientation of common words. Finally, the words＇ semantic orientation and the context were combined to recognize the given words＇ orientation. The experiments show that the measurement approach achieves better results for common words＇ orientation classification and contributes particularly to the text orientation classification of large granularities.展开更多
基金supported by the National Natural Science Foundation of China (50375010).
文摘Based on the text orientation classification, a new measurement approach to semantic orientation of words was proposed. According to the integrated and detailed definition of words in HowNet, seed sets including the words with intense orientations were built up. The orientation similarity between the seed words and the given word was then calculated using the sentiment weight priority to recognize the semantic orientation of common words. Finally, the words＇ semantic orientation and the context were combined to recognize the given words＇ orientation. The experiments show that the measurement approach achieves better results for common words＇ orientation classification and contributes particularly to the text orientation classification of large granularities.
