A Taxonomy of Programming Languages for Code Generation
This provides a framework for dataset curation and evaluation of multilingual LLMs, addressing a gap in code generation research, though it is incremental as it adapts an existing taxonomy concept from natural languages.
The authors tackled the lack of a resource-tier taxonomy for programming languages in code generation by creating the first reproducible classification of 646 languages into four tiers, revealing that only 1.9% of languages account for 74.6% of tokens in major corpora while 71.7% contribute just 1.0%.
The world's 7,000+ languages vary widely in the availability of resources for NLP, motivating efforts to systematically categorize them by their degree of resourcefulness (Joshi et al., 2020). A similar disparity exists among programming languages (PLs); however, no resource-tier taxonomy has been established for code. As large language models (LLMs) grow increasingly capable of generating code, such a taxonomy becomes essential. To fill this gap, we present the first reproducible PL resource classification, grouping 646 languages into four tiers. We show that only 1.9% of languages (Tier 3, High) account for 74.6% of all tokens in seven major corpora, while 71.7% of languages (Tier 0, Scarce) contribute just 1.0%. Statistical analyses of within-tier inequality, dispersion, and distributional skew confirm that this imbalance is both extreme and systematic. Our results provide a principled framework for dataset curation and tier-aware evaluation of multilingual LLMs.