Identifying semantically equivalent multi-word expressions (phrases) is one of the core challenges in natural language processing, especially for low-resource languages such as Uzbek. This study explores the practical application of the GloVe algorithm in identifying synonymous phrases in the Uzbek language. A specialized dataset containing 298 phrases and their synonymous counterparts was compiled for the research. Based on this dataset, pre-trained 100-dimensional GloVe vectors were employed to represent phrases using the average vectorization method. Semantic similarity between phrases was evaluated using cosine similarity, and a threshold-based classification approach was applied to determine synonymy. The experimental results demonstrated that the GloVe model shows a certain level of effectiveness in capturing semantic similarity among Uzbek phrases.