This paper examines semantic similarity and intertextuality in selected texts from the Vedic Sanskrit corpus, specifically the Maitrāyaṇī Saṃhitā (MS) and Kāṭhaka-Saṃhitā (KS). Three computational methods are employed: Word2Vec for word embeddings, stylo package for stylometric analysis, and TRACER for text reuse detection. By comparing various sections of the texts at different granularities, patterns of similarity and structural alignment are uncovered, providing insights into textual relationships and chronology. Word embeddings capture semantic similarities, while stylometric analysis reveals clusters and components that differentiate the texts. TRACER identifies parallel passages, indicating probable instances of text reuse. The computational analysis corroborates previous philological studies, suggesting a shared period of composition between MS.1.9 and MS.1.7. This research highlights the potential of computational methods in studying ancient Sanskrit literature, complementing traditional approaches. The agreement among the methods strengthens the validity of the findings, and the visualizations offer a nuanced understanding of textual connections. The study demonstrates that smaller chunk sizes are more effective for detecting intertextual parallels, showcasing the power of these techniques in unraveling the complexities of ancient texts.

Neural Machine Translation (NMT) has revolutionized language translation, yet significant challenges persist for low-resource languages, particularly those with high dialectal variation and limited standardization. This comprehensive study focuses on the Ainu language, a critically endangered indigenous language of northern Japan, which epitomizes these challenges. We address the limitations of previous research through two primary strategies: (1) extensive corpus expansion encompassing diverse domains and dialects, and (2) development of innovative methods to incorporate dialect and domain information directly into the translation process. Our approach yielded substantial improvements in translation quality, with BLEU scores increasing from 32.90 to 39.06 (+6.16) for Japanese → Ainu and from 10.45 to 31.83 (+21.38) for Ainu → Japanese. Through rigorous experimentation and analysis, we demonstrate the crucial importance of integrating linguistic variation information in NMT systems for languages characterized by high diversity and limited resources. Our findings have broad implications for improving machine translation for other low-resource languages, potentially advancing preservation and revitalization efforts for endangered languages worldwide.

The advent of Large Language Models (LLMs) substantially raised the quality and lowered the cost of Machine Translation (MT). Can scholars working with ancient languages draw benefits from this new technology? More specifically, can current MT facilitate multilingual digital papyrology? To answer this question, we evaluate 9 LLMs in the task of MT with 4 Coptic and 4 Ancient Greek ostraca into English using 6 NLP metrics. We argue that some models have already reached a performance apt to assist human experts. As can be expected from the difference in training corpus size, all models seem to perform better with Ancient Greek than with Coptic, where hallucinations are markedly more common. In the Coptic texts, the specialised Coptic Translator (CT) competes closely with Claude 3 Opus for the rank of most promising tool, while Claude 3 Opus and GPT-4o compete for the same position in the Ancient Greek texts. We argue that MT now substantially heightens the incentive to work on multilingual corpora. This could have a positive and long-lasting effect on Classics and Egyptology and help reduce the historical bias in translation availability. In closing, we reflect upon the need to meet AI-generated translations with an adequate critical stance.

LAJaR (Language Atlas of Japanese and Ryukyuan) is a linguistic typology database focusing on micro-variation of the Japonic (Japanese and Ryukyuan) languages. This paper aims to report the design and progress of this ongoing database project. Finally, we also show a case study utilizing its database on zero copulas among the Japonic languages.

The Open Multilingual Online Lexicon of Okinawan (OMOLO) project aims to create an accessible, user-friendly digital lexicon for the endangered Okinawan language using digital humanities tools and methodologies. The multilingual web application, available in Japanese, English, Portuguese, and Spanish, will benefit language learners, researchers, and the Okinawan community in Japan and diaspora countries such as the U.S., Brazil, and Peru. The project also lays the foundation for an Okinawan UD Treebank, which will support computational analysis and the development of language technology tools such as parsers, machine translation systems, and speech recognition software. The OMOLO project demonstrates the potential of computational linguistics in preserving and revitalizing endangered languages and can serve as a blueprint for similar initiatives.

This paper explores the potential of Machine Translation (MT) in preserving and revitalizing Ainu, an indigenous language of Japan classified as critically endangered by UNESCO. Through leveraging Marian MT, an open-source Neural Machine Translation framework, this study addresses the challenging linguistic features of Ainu and the limitations of available resources. The research implemented a meticulous methodology involving rigorous preprocessing of data, prudent training of the model, and robust evaluation using the SacreBLEU metric. The findings underscore the system’s efficacy, achieving a SacreBLEU score of 32.90 for Japanese to Ainu translation. This promising result highlights the capacity of MT systems to support language preservation and aligns with recent research emphasizing the potential of computational techniques for low-resource languages. The paper concludes by affirming the significant role of MT in the broader context of language preservation, serving as a crucial tool in the fight against language extinction. The study paves the way for future research to harness advanced MT techniques and develop more sophisticated models for endangered languages.

With the increasing availability of wordnets for ancient languages, such as Ancient Greek and Latin, gaps remain in the coverage of less studied languages of antiquity. This paper reports on the construction and evaluation of a new wordnet for Coptic, the language of Late Roman, Byzantine and Early Islamic Egypt in the first millenium CE. We present our approach to constructing the wordnet which uses multilingual Coptic dictionaries and wordnets for five different languages. We further discuss the results of this effort and outline our on-going/future work.