<p>Cross-lingual sentiment analysis is a challenging task in natural language processing that aims to analyze and understand sentiments expressed in texts across different languages. The motivation behind this task is to address the issue faced by sentiment analysis, where the majority of languages lack sufficient labeled data for training. Nevertheless, cross-lingual sentiment analysis tasks mostly depend on translation tools and corpora. This dependency can prevent these models from capturing the unique sentiment characteristics in the target language due to the accuracy limitations of translation tools and corpora. Driven by the common belief that emojis have consistent associations with sentiments across languages, we introduce a method called emotional mutual reinforcement (EMR). Our EMR approach aims to transfer sentiment knowledge between different languages by using emojis as medium of connection. The central concept of EMR involves employing emojis as the medium to connect different languages, facilitating emotional mutual reinforcement across different linguistic contexts. Sentiment knowledge can effectively cross linguistic boundaries through reinforcing emotion-related features with similar sentiment tendencies. In contrast to existing approaches that utilize emojis as a bridge, EMR can incorporate more fine-grained sentiment knowledge from different languages. Results from a complete evaluation on several publicly available datasets confirm the efficiency of the approach we proposed.</p>