The study proposes a methodology for constructing emotion-sentiment profiles of customer reviews based on cluster analysis. The approach employs k-means clustering to group reviews into distinct emotion clusters. For each review, a multidimensional vector representation is generated by calculating centroid vectors of emotion-related parameters (valence, arousal, and five discrete emotions) across sentiment categories (positive, negative, their intense variants, and also neutral). The emotion parameters are derived from the lexical database ENRuN-2, where words are annotated with mean and standard deviation scores for emotion dimensions. The sentiment category reference is derived from the corpus annotation capturing sentiment related phrases in the text. The generated text profile integrates both global text characteristics and localized patterns within sentiment-marked phrases, facilitating correlation analysis between emotion and sentiment categories. Using a kNN algorithm, a test review is mapped to the relevant document cluster. Further, aggregated emotion scores from the closest neighboring datapoints within the cluster are calculated for each sentiment category individually, thereby the emotional load of the review under analysis is identified. The proposed approach was validated on a corpus of healthcare service reviews, comprising more than three thousand reviews of 200,000 tokens in total. The corpus was divided into sentiment-annotated training and unannotated test splits. The results highlight the utility of cluster-based profiling in maintaining interpretability through dimensional approach to emotion and sentiment-related text representation, while enabling scalable analysis of customer feedback within the chosen domain.