🤖 AI Summary
This study addresses the challenge of inferring language genealogical relationships solely from transcription-based textual fragments—without relying on translation or decipherment—particularly for extant languages, deciphered ancient languages, and the undeciphered Cypro-Minoan script. Method: We propose the first end-to-end framework leveraging convolutional generative adversarial networks (c-GANs) for language phylogeny modeling, integrating transcriptional text representations with unsupervised and weakly supervised representation learning to bypass traditional comparative linguistics’ dependence on cognate identification and manually constructed sound-change rules. Contribution/Results: Our approach successfully reconstructs established phylogenetic structures across multiple ancient languages and yields the first deep learning–based genealogical attribution for Cypro-Minoan, suggesting its plausible affiliation with the Aegean/Anatolian linguistic sphere. The method is both interpretable and scalable, offering a novel paradigm for phylogenetic placement and subsequent decipherment of undeciphered scripts.
📝 Abstract
We use a c-GAN (convolutional generative adversarial) neural network to analyze transliterated text fragments of extant, dead comprehensible, and one dead non-deciphered (Cypro-Minoan) language to establish linguistic affinities. The paper is agnostic with respect to translation and/or deciphering. However, there is hope that the proposed approach can be useful for decipherment with more sophisticated neural network techniques.