Abstract:

An automatic scoring method based on evaluation content is proposed. The shortcoming of traditional words similarity computing methods is insufficient in sentiment orientation judgement is analyzed, and hence a solution method is proposed. The impact of sentiment modifier on sentiment orientation is discussed, and standard words are selected with the help of UNICET,which is a simulated software in social network. The experimental results show that the proposed automatic scoring method is more approximate to manual scoring, thus having certain reasonableness.

Key words: user evaluation;automatic scoring;sentiment orientation;semantic similarity;sentiment modifier