Abstract: With the advent of the big data era， the scale of e-commerce platforms has expanded rapidly and consumers need more efficient product selection methods， which would help make decisions quickly in the massive kinds and a number of products. Therefore， a method of product selection based on online reviews is proposed. First， the utility values of the online reviews are calculated. By extracting the product attributes， the attribute set for product selection is obtained. Considering the utility values of the online reviews， the weight vector of attributes is determined. Second， by analyzing the sentiment words， consumers’ sentiment preferences are expressed in the format of probability distribution about sentiment levels. Based on the principle of stochastic dominance， the dominance relationships of any pairwise products on each attribute are determined. Third， by using PROMETHEE II， the ranking result of alternative products is obtained. Finally， an example of product selection is given to illustrate the feasibility and practicability of the proposed method.

Key words: product selection, online review, attribute extraction, sentiment analysis, stochastic dominance