Asian Journal of Mathematics and Computer Research

Kumaoni culture, which finds its origin in the Indian state of Uttarakhand, is richly endowed with traditions, language, festivals, and arts. However, with modernization, migration, and globalization, it has slowly started to die out. In this work, we present a complete computational analysis for the cultural decline of Kumaon through methods of Natural Language Processing and Machine Learning techniques. A multilingual corpus was built comprising historical texts, digital media, and oral traditions. The sentiments that arose from the digital texts in the last decade reveal the negative trend of almost 40% in the decline of references to important cultural elements like Harela and Aipan art. A predictive model helps in identifying some of the traditions at the highest risk so that they are preserved immediately. The parallel English-Kumaoni data was fed into neural machine translation systems like mBART and IndicBERT, thus achieving best performance around epoch 3 with a training loss of 0.2961. The study also propounds the idea for AI-powered translation platforms, VR experiences, and learning tools for young generation engagement and cultural rejuvenation. These findings have a strong bearing on policy advocacy, curricular efforts, and digital preservation. Thus, this interdisciplinary work contributes toward.