We present a dataset for Symbol-level language modelling of linear B scripts using a Bidirectional Recurrent Neural Network. The RNN is trained to predict the missing symbol in a given Mycenaean sequence. It includes Series "D" sequences 2565 sequences without missing symbols extracted from: J. Chadwick, L. Godart, J. T. Killen, J. P. Olivier, A. Sacconi, and I. A. Sakellarakis. Corpus of Mycenaean Inscriptions from Knossos: Volumes 1-4, Cambridge University Press, 1987. The first 513 sequences result from the processing of series D tablets, The next 725 sequences result from the augmentation techniques described in the paper The remaining 1327 sequences are duplicates to minimize bias. The dataset was used in the following publication: K. PAPAVASSILEIOU, Kosmopoulos, D. I., G. Owens, “A generative model for the Mycenaean Linear B script and its application in infilling text from ancient tablets”, ACM Journal on Computing and Cultural Heritage, https://doi.org/10.1145/3593431. If you use this dataset please cite this publication.