With opinionated text becoming an overabundant resource due to the Web, the need for processing this resource is becoming more and more relevant. That is why the field of Aspect-Based Sentiment Classification (ABSC) has seen rapid development. However, most research has been focused on English texts. Therefore, this paper adds to the field of Multilingual Aspect-Based Sentiment Classification (MABSC) and Cross-lingual Aspect-Based Sentiment Classification (XABSC). We take a state-of-the-art ABSC model, LCR-Rot-hop++, and repurpose it for MABSC and XABSC. We propose MABSC models mLCR-Rot-hop++ and MLCR-Rot-hop++, which use mBERT and a multilingual dataset, respectively. For XABSC we propose MLCR-Rot-hop-XX \(_\text {en}\) , which uses translation techniques from English to another language (XX) to form the training data. Furthermore, we make use of Aspect-Code-Switching (ACS) to further extend the training data and make it bilingual. Last, we also introduce Unilingual ABSC (UABSC) models, which are models trained on resource-poor languages. These models are called mLCR-Rot-hop-XX++. The best performance for MABSC is shown by MLCR-Rot-hop++. Furthermore, mLCR-Rot-hop++ is our best model for XABSC. The UABSC models mLCR-Rot-hop-XX++ are the best performing overall but these are trained and tested on resource-poor languages.