New framework BERTilda tracks topic evolution in text streams

Researchers have developed BERTilda, a new framework designed to track the evolution of topics within longitudinal text streams. This system identifies topic birth, death, splits, and merges by constructing a temporal topic graph. BERTilda uses both semantic similarity and document outflow/inflow signals to link topics across time windows, achieving high agreement rates on annotated datasets for lifecycle labeling. AI

IMPACT Provides a novel method for analyzing the lifecycle of topics in large text datasets, potentially improving information retrieval and trend analysis.

RANK_REASON Academic paper detailing a new framework for topic modeling. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CL →

• arXiv
• Hugging Face
• U.S.

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →