Concept graph unlocks Swami Vivekananda’s texts
Summary and headline written by AI from the source article. How we work
Researchers developed Viveka-Insight, a new resource for exploring the complete works of Swami Vivekananda, a 19th-century Indian philosopher and spiritual leader.
The project combines his writings in English and Bengali, totaling around 15 million characters, into a linked database of concepts and citations. Viveka-Insight includes a detailed breakdown of the texts into over 32,000 paragraphs and 168,000 sentences, alongside a concept graph connecting 8,362 ideas across both languages.
This graph uses English labels to link related passages even without direct translations. The team verified the system’s ability to find information across languages, achieving 86% recall when searching for known lecture pairs in both English and Bengali. They also assessed the accuracy of concept extraction, finding a precision of 0.60 when multiple annotators agreed. While concept extraction performed better in English than Bengali, the system demonstrated potential for application to other collections of classical texts in multiple languages.
The resource and pipeline are openly available.
