<![CDATA[The Kudus regional language is an important cultural identity, but it is now facing extinction due to its lack of use by the younger generation, the popularity of slang and international languages, and the lack of digital text resources. General machine translation tecnologies such as Google translate still have limitations in understanding specific cultural contexts and lokal dialects. The main objective is to develop an automatic translation tool from Indonesian to the Kudus regional language. Utilizing RNN-GRU-based Neural Machine Translation (NMT) technology to overcome translation limitations in the low resource language category. Through an experimental quantitative approach by applying the NMT model of the Recurent Neural Network architecture, specifically Gated Recurent Unit (RNN-GRU). Using data in the frm of a parallel corpus with a total 1,031 pairs of sentences that have been cleaned and tokenized, the divided into three parts 824 for training, 103 for validation, and 104 for testing. The model was built using Tensorflow and trained for a maximum of 106 epochs with early stopping, Adam optimizer, and bacth 64 on Colab GPU. The developed model achieved a BLEU score of 0.89% on the testing data, demonstrating the significant complexity and challenges of translating low-resource regional dialects. These findings prove that the RNN-GRU model is further development is still needes by expanding the number of parallel corpora and exploring more state-of-the-art architectures to improve vocabulary richness and accuracy]]>
