Internet has become the source of almost everything, ranging from voluminous information to products and services. Sorting through the countless comments may be a waste of time on the part of the customers and on manufacturer’s human resources. We aim to mine the important information of product review from Internet customers. We propose an effective way to extract product features and opinion words simultaneously from cosmetic customer reviews in Chinese. The process consists of two parts. First, several seed words are used to generate pattern rules; second, pattern rules, iterative process, and negation words identification methods are employed to automatically extract opinion units that commented by users. Our experimental results indicate that our proposed techniques are highly effective.