Mitrasamgraha: A Comprehensive Classical Sanskrit Machine Translation Dataset

TL;DR

Mitrasaṃgraha tackles the scarcity and narrow scope of Sanskrit–English MT data by introducing the largest public parallel corpus to date (391,548 sentence pairs) spanning Vedic to late-medieval works across six domains. The authors compare alignment strategies (BertAlign favored) and provide manually verified development/test splits, with metadata enabling fine-grained domain/time analysis. They benchmark commercial and open models, showing that neural metrics like BLEURT and GEMBA correlate best with human judgments and that fine-tuning and retrieval-augmented generation improve translation quality. The dataset supports rigorous MT research and digital humanities in Sanskrit, while also highlighting persistent linguistic challenges such as sandhi, compounding, and metaphoric language that limit current systems.

Abstract

While machine translation is regarded as a "solved problem" for many high-resource languages, close analysis quickly reveals that this is not the case for content that shows challenges such as poetic language, philosophical concepts, multi-layered metaphorical expressions, and more. Sanskrit literature is a prime example of this, as it combines a large number of such challenges in addition to inherent linguistic features like sandhi, compounding, and heavy morphology, which further complicate NLP downstream tasks. It spans multiple millennia of text production time as well as a large breadth of different domains, ranging from ritual formulas via epic narratives, philosophical treatises, poetic verses up to scientific material. As of now, there is a strong lack of publicly available resources that cover these different domains and temporal layers of Sanskrit. We therefore introduce Mitrasamgraha, a high-quality Sanskrit-to-English machine translation dataset consisting of 391,548 bitext pairs, more than four times larger than the largest previously available Sanskrit dataset Itih=asa. It covers a time period of more than three millennia and a broad range of historical Sanskrit domains. In contrast to web-crawled datasets, the temporal and domain annotation of this dataset enables fine-grained study of domain and time period effects on MT performance. We also release a validation set consisting of 5,587 and a test set consisting of 5,552 post-corrected bitext pairs. We conduct experiments benchmarking commercial and open models on this dataset and fine-tune NLLB and Gemma models on the dataset, showing significant improvements, while still recognizing significant challenges in the translation of complex compounds, philosophical concepts, and multi-layered metaphors. We also analyze how in-context learning on this dataset impacts the performance of commercial models

Figures (3)

• Figure 1: Flowchart of our Sanskrit-English bitext collection and alignment procedure.
• Figure 2: Distribution of Sanskrit Literature Categories
• Figure 3: Average Pearson (blue) and Spearman (red) correlations of automatic metrics vs. human judgments. GEMBA* is the reference-free implementation of GEMBA gemba, while GEMBA is reference-based.