INTSUMGen: Structured INTSUM Dataset for Defense and Security Intelligence Extraction
收藏DataCite Commons2026-05-02 更新2026-05-07 收录
下载链接：
https://zenodo.org/doi/10.5281/zenodo.19950394
下载链接
链接失效反馈官方服务：
资源简介：
Structured INTSUM Dataset for Military, Conflict, and Security Intelligence Extraction
This dataset contains structured Intelligence Summary (INTSUM) reports generated from real-world defense, military, and conflict-related news sources.
Each entry follows a standardized INTSUM format used in intelligence analysis, including fields such as:- Date- Location - Event Type- Actors Involved (Primary and Secondary)- Event Summary- Outcome- Strategic Implications- Recommended Actions
The dataset is constructed using a Large Language Model-based structured extraction (Mistral/OpenRouter)
It includes both:- Real-world extracted intelligence summaries- Synthetic event-specific INTSUM samples for model training and benchmarking
This dataset is designed for use in:- Natural Language Processing (NLP)- Information Extraction- Retrieval-Augmented Generation (RAG)- Defense and security analytics- Event classification and summarization- Intelligence automation systems
The dataset can support training and evaluation of models for structured intelligence generation, entity extraction, and decision-support systems in security domains.
Format:- JSONL (for model training and RAG pipelines)- CSV (synthetic structured events)
提供机构：
Zenodo
创建时间：
2026-05-02
