In this paper, we propose a novel sentiment-aware topic model to construct a topic-adaptive sentiment lexicon (TaSL) for higher-level sentiment classification tasks. It's widely recognized that sentiment lexicon as a useful prior knowledge is crucial in sentiment analysis. However, many existing sentiment lexicons are constructed ignoring the variability of the sentiment polarities of words in different topics or domains. In this paper, we address this issue by assigning multiple pairs of topics and sentiments for each word. In TaSL, documents are represented by multiple pairs of topics and sentiments and words are characterized by a multinomial distribution over the pairs of topics and sentiments. The main advantage of TaSL is that the sentiment polarities of words in different topics can be sufficiently captured. This model is beneficial to construct a topic-specific sentiment lexicon and then effectively improve the performance of sentiment classification. Extensive experimental results on four publicly available datasets RM, OMD, semEval13A and semEval16B were presented to demonstrate the usefulness of the proposed approach. The results show that TaSL performs better than the state-of-the-art methods.