Discover ideas, stories, and research shaping the future of multilingual AI. From field updates to behind-the-scenes reflections, explore what we're building, learning, and sharing across Zoorna.
Timeline generation is the process of automatically identifying, ordering, and anchoring events in time based on information in text. This post illustrates the challenges in identifying and resolving temporal expressions for automatic timeline generation in Persian text.
Zoorna Institute presented new research at NACIL4 on using large language models (LLMs) for narrative analysis in Persian and Sorani Kurdish—two languages often excluded from mainstream AI. Our approach leverages prompt-based LLM workflows to extract events, timelines, and implicit meaning without building full NLP pipelines. The results highlight both the potential and limitations of LLMs in low-resource language contexts.
