With the development of the e-commerce market, customers have faced an information overload problem. Therefore, the importance of personalized recommendation becomes more important because it helps to reduce the cost of decision-making. Recently, many studies on online review-based recommender systems have been actively conducted, extracting consumers’ preference information from reviews. Specifically, most studies only use polarity factors of reviews as consumers’ preference information. However, besides polarity factors, various emotions still exist in the review that can be seen as essential consumer features. Therefore, this study proposed a novel recommendation model that applies multidimensional emotional factors of reviews to address such a limitation. For this, this study applies Multi-layer Perceptron to learn nonlinear consumer-product interaction. Then, this study extracts eight emotional factors from online reviews by combining the NRC dictionary and the LIWC program. Finally, this study predicts consumer preference based on the consumer-product interaction vector and eight emotional factors. To evaluate recommendation performance, this study used the movie review dataset collected from Amazon.com, which is one of the most representative experience goods. The experimental results showed the proposed model outperforms the benchmark models. The core reason is that the proposed model effectively used consumer features with eight emotional factors in reviews. Therefore, the proposed model in this study can provide an enhanced recommendation service by using online reviews