With the development of e-commerce, the national e-commerce transaction volume sprung up. An increasing number of customers have made it a habit to contact online customer service when they run into difficulties while shopping online. These conversation texts include a large number of emotional words and tone particles, which can intuitively reflect customers' attitudes toward any situation, product, or service. The intelligent customer service conversation texts have the characteristics of serious colloquialism, high diversity of words, and short text length. Traditional sentiment analysis algorithms are not suitable for dialogue information. Besides, coarse-grained sentiment analysis could not fully display the text information of users in the dialogue, resulting in poor performance of sentiment classification prediction. In this paper, we propose a hybrid word embedding method based on Gaussian distribution to leverage the emotional syntactic and semantic richness of the two distributed word representations. Furthermore, this study utilizes a stacked ensemble method by combining the outputs obtained from three deep learning models (i.e., CNN, LSTM, and GRU) for simultaneously predicting coarse-grained and fine-grained sentiment analysis in the customer service conversation domain. The results show that hybrid word embedding (HWE) can assist us in comprehending our word representations in context more effectively. In comparison to the current state-of-the-art models, our proposed ensemble model significantly enhances sentiment classification performance.