The task of identifying the sentiment polarity of terms is to decide whether an individual term is subjective or objective, and to classify a subjective term as positive or negative. Accurate identification of sentiment polarity of terms is of vital importance in sentiment analysis, as well as a challenging task. As a specific semantic subspace, S-HAL (Sentiment Hyperspace Analogue to Language) shows the advantages in modelling and distinguishing semantic orientation characteristics of terms. In this article, we present a new method to identify the sentiment polarity of Chinese words, which is based on S-HAL model and standard supervised learner. A series of empirical evaluation results demonstrate that S-HAL-based identification method could outperform the known method and the way of combining multiple classifiers can balance the identification performance between subjective and objective. Similar to the way for building SentiWordNet from WordNet, the method presented in this article provides a solid technical basis for construction of Chinese SentiHowNet based on HowNet.