In the fast moving world, peoples want to reduce the shopping time and purchase their needed products through online. In addition, online shopping provides the product reviews and helps the customer to get the better among the variety of brands. In this, mining the opinion words and the polarity of the reviews are the important task to detect the exact opinion of the customer reviews. In this paper, a novel approach is proposed based on the semi-supervised word alignment model (SWAM), which identifies the relations among the words in a sentence. It's a graph-based algorithm where target opinion words are compared with the other opinion word and extract the long span relations among the words. Unlike, syntax based method, this proposed model reduce the parsing errors by dealing with informal online texts. The mined reviews of this proposed system provides better precision when compared with standard unsupervised alignment review models. The experimental results show that this approach effectively mining the user reviews and provide the better recommendation. © 2017 IEEE.