• Digital Humanities and Cultural Heritage
• Open Access close
• Publications close
• Research software close
• Other research products close
• JP close
• KR close

This study presents a groundbreaking approach to preserving the Ainu language, recognized as critically endangered by UNESCO, by developing a bi-directional neural machine translation (MT) system between Ainu and Japanese. Utilizing the Marian MT framework, known for its effectiveness with resource-scarce languages, the research aims to overcome the linguistic complexities inherent in Ainu's polysynthetic structure. The paper delineates a comprehensive methodology encompassing data collection from diverse Ainu text sources, meticulous preprocessing, and the deployment of neural MT models, culminating in the achievement of significant SacreBLEU scores that underscore the models' translation accuracy. The findings illustrate the potential of advanced MT technology to facilitate linguistic preservation and educational endeavors, advocating for integrating such technologies in safeguarding endangered languages. This research not only underscores the critical role of MT in bridging language divides but also sets a precedent for employing computational linguistics to preserve cultural and linguistic heritage.

I analyze factory worker households in the early 1920s in Osaka to examine idiosyncratic income shocks and consumption. Using the household-level monthly panel dataset, I find that while households could not fully cope with idiosyncratic income shocks at that time, they mitigated fluctuations in indispensable consumption during economic hardship. In terms of risk-coping mechanisms, I find suggestive evidence that savings institutions helped mitigate vulnerabilities and that both using borrowing institutions and adjusting labor supply served as risk-coping strategies among households with less savings.

The weak correlation between spatiotemporal fluctuations in nonequilibrium complex systems is shown to govern the fluctuation distribution, maximizing the conditional entropy associated with such fluctuations. The result is illustrated in diffusion phenomena observed in living cells. A generic feature of the weak correlation is briefly mentioned. Comment: 16 pages, no figures. For the proceedings of the 12th International Conference on Mathematical Modeling in Physical Sciences (August 28-31, 2023, Belgrade, Serbia)

Sake (Japanese rice wine) was fermented in pottery for more than a millennium before wooden barrels were adopted to obtain a greater brewing capacity. Although a recently conducted analysis of sake brewed in pottery indicated that sake brewed in unglazed pottery contains more ethanol than that brewed in glazed pottery, little is known about the characteristics of sake brewed in pottery. In this study, we used two types of ceramic containers of identical size, one glazed and one unglazed, for small-scale sake brewing to evaluate the effects of glazing on fermentation properties. The following parameters were measured continuously in the sake samples over 3 weeks of fermentation: temperature, weight, ethanol concentration, and glucose concentration in sake mash. Taste-sensory values, minerals, and volatile components were also quantified in the final fermented sake mash. The results show that, in the unglazed containers, the temperature of the sake mash was lower and the weight loss was higher compared to the sake mash in the glazed containers. The quantity of ethanol and the levels of Na+, Fe3+, and Al3+ tended to be higher in the sake brewed in the unglazed pottery. A taste-sensory analysis revealed that umami and saltiness were also higher in the samples brewed in the unglazed pottery. These results suggest that glazing affects multiple fermentation parameters and the flavor of sake brewed in pottery. They may also suggest that the materials of the containers used in sake brewing generally affect the fermentation properties.

This research investigates consumer reviews of eco-friendly products on Amazon to uncover valuable sustainability insights that can inform design optimization. Using natural language processing (NLP) techniques, including sentiment analysis, key terms extraction, and topic modeling, this research reveals diverse perspectives related to sustainability aspects in eco-friendly products. Innovatively, we integrate the NLP approach with correspondence analysis (CA) to understand consumer sentiments and preferences related to sustainability aspects. Leveraging CA, we visualize the interplay between eco-friendly product features and consumer sentiments, revealing underlying relationships and patterns. The CA biplot showcases the alignment of specific sustainability attributes with consumer satisfaction, highlighting which sustainability aspects hold greater influence over overall product ratings. As sustainability becomes an increasingly crucial aspect of consumer choices, our paper emphasizes the significance of a multidimensional approach that embraces both qualitative and quantitative insights. By blending CA with consumer reviews, we equip designers and stakeholders with an innovative and comprehensive toolkit to enhance sustainable design practices, paving the way for more informed and effective product development strategies in the realm of eco-friendliness.

International audience; This paper presents a series of studies on the impact of regional variation in L1 (source language) on L2 (target language) pronunciation, focusing on the high back vowel /u/ in French perceived or produced by native speakers of Japanese from the Kanto area (around Tokyo) and the Kansai area (around Osaka). 1) To serve as a base line, the 5 Japanese vowels were pronounced by 11 speakers (5 women and 6 men) from Kansai with no learning experience of French and their first 4 formants were measured to examine their acoustic properties. 2) 25 naïve listeners from Kansai took part in an AXB auditory discrimination task for French vowels including pairs /u/-/y/, /y/-/ø/ and /u/-/ø/; the results were compared with those of Kanto speakers in previous studies. 3) Some Japanese-speaking learners of French (JSL) from Kansai produced the French /u/ with a lower F2 (around 1,000 Hz for men) than typical values observed for learners from the Tokyo area. These results show limited differences for some of the speakers from Kansai compared to the tendency commonly observed with Kanto speakers.; Cet article présente une série d’études portant sur l’impact de la variation régionale en L1 (langue source) sur la prononciation en L2 (langue cible), en se focalisant sur la voyelle fermée postérieure /u/ en français perçue ou produite par des locuteurs natifs du japonais du Kanto (région de Tokyo) et du Kansaï (région d’Osaka). 1) Pour servir de référence, les 5 voyelles du japonais ont été prononcées par 11 locuteurs (5 femmes et 6 hommes) du Kansaï (Osaka) sans expérience d’apprentissage du français et les 4 premiers formants ont été mesurés afin d’examiner leurs propriétés acoustiques. 2) 25 auditeurs originaires du Kansaï et non-apprenants du français ont effectué une tâche de discrimination auditive AXB de voyelles françaises, comprenant les paires /u/-/y/, /y/-/ø/ et /u/-/ø/ : les résultats ont été comparés avec ceux de locuteurs du Kanto dans des études antérieures. 3) Certains apprenants japonophones du français langue étrangère du Kansaï ont produit le /u/ français avec un F2 plus bas (autour de 1000 Hz pour les hommes) que les valeurs typiques observées chez des apprenants du Kanto. Ces résultats montrent des différences limitées chez certains des locuteurs du Kansaï par rapport à la tendance communément observée chez des locuteurs du Kanto.

Despite a few attempts to automatically crawl Ewe text from online news portals and magazines, the African Ewe language remains underdeveloped despite its rich morphology and complex "unique" structure. This is due to the poor quality, unbalanced, and religious-based nature of the crawled Ewe texts, thus making it challenging to preprocess and perform any NLP task with current transformer-based language models. In this study, we present a well-preprocessed Ewe dataset for low-resource text classification to the research community. Additionally, we have developed an Ewe-based word embedding to leverage the low-resource semantic representation. Finally, we have fine-tuned seven transformer-based models, namely BERT-based (cased and uncased), DistilBERT-based (cased and uncased), RoBERTa, DistilRoBERTa, and DeBERTa, using the preprocessed Ewe dataset that we have proposed. Extensive experiments indicate that the fine-tuned BERT-base-cased model outperforms all baseline models with an accuracy of 0.972, precision of 0.969, recall of 0.970, loss score of 0.021, and an F1-score of 0.970. This performance demonstrates the model’s ability to comprehend the low-resourced Ewe semantic representation compared to all other models, thus setting the fine-tuned BERT-based model as the benchmark for the proposed Ewe dataset.