Data and code for "Context-Biased Entity Summarization for Video Enrichment".
收藏官方服务：
资源简介：
Data and code supporting the paper “Context-Biased Entity Summarization for Video Enrichment: Selecting DBpedia Facts by Video Context and Evaluating Them with Humans and an LLM Judge”. The method ranks the DBpedia facts of a linked entity by their similarity to the video context and gives only the top facts to a language model that writes a short caption. The package contains the 37-entity benchmark, the generated captions and LLM-judge scores, the ratings of two human annotators, and the code.
提供机构：
Zenodo创建时间：
2026-11-04
