Fujitsu/ForQA_Knowledge_Dataset

# ForQA_Knowledge_Dataset

## Dataset Summary
- This dataset is a structured knowledge graph constructed using Fujitsu Knowledge Graph enhanced RAG for Q&A 2.0, one of Fujitsu's proprietary technology.
- `Fujitsu-ShareholdersMTG-FY2023.pdf` (Fujitsu Group Business Report FY2023 (Provisional)*) was translated to English, and English document was used as the input document and converted into a knowledge graph.
- Because it is provided in Neo4j format, it can be easily visualized and can be utilized for simple chat applications based on the knowledge graph, as well as for research and development of AI technologies.

## Repository Overview
| Name | Description |
| ----------- | ----------- |
| `forqa2_en.graphml` | Knowledge Graph for Q&A 2.0 constructed from `Fujitsu-ShareholdersMTG-FY2023.pdf` |
| `Fujitsu-ShareholdersMTG-FY2023.pdf` | Fujitsu Group Business Report FY2023 (Provisional)* |

* Note: The title provided is a provisional name for analysis purposes and may differ from the official document title.

## How to Use the Data
Demo code for using this data is available at:
https://github.com/FujitsuResearch/Knowledge_Data/

## Publication
Taku Fukui and Satoshi Munakata, "画像を含む文書から検索用洞察を生成することによるマルチモーダルRAGシステムの検索精度の改善", In Proceedings of the 39th Annual Conference of the Japanese Society for Artificial Intelligence (JSAI2025), 4Q1-GS-10-01, 2025.

Kenichirou Narita and Satoshi Munakata, "Chunk-Link: Context-aware chunk completion", In Proceedings of the 2nd Workshop on Retrieval-Augmented Generation over Knowledge Graphs (RAGE-KG 2025), CEUR Workshop Proceedings, Vol-4079, short1, 2025.

## License
CC-BY-4.0

## Contacts
Kenichirou Narita: k.narita{at}fujitsu.com

# ForQA_Knowledge_Dataset
## 数据集概览
- 本数据集是依托富士通（Fujitsu）专有技术之一的增强型检索增强生成（Retrieval-Augmented Generation，RAG）问答2.0（Q&A 2.0）系统构建的结构化知识图谱。
- 本数据集以`Fujitsu-ShareholdersMTG-FY2023.pdf`（即《富士通集团2023财年业务报告（暂定版）》*）的英文译稿作为输入文档，转换为知识图谱。
- 由于该数据集以Neo4j格式提供，因此可便捷实现可视化，既可用于基于知识图谱的简易聊天应用开发，也可用于人工智能技术的研发工作。

## 仓库概览
| 名称 | 描述 |
| ----------- | ----------- |
| `forqa2_en.graphml` | 由`Fujitsu-ShareholdersMTG-FY2023.pdf`构建的问答2.0专用知识图谱 |
| `Fujitsu-ShareholdersMTG-FY2023.pdf` | 富士通集团2023财年业务报告（暂定版）* |

* 注：本次提供的文档标题为用于分析的暂定名称，可能与官方文档正式标题存在差异。

## 数据使用指南
本数据集的演示代码可通过以下链接获取：
https://github.com/FujitsuResearch/Knowledge_Data/

## 发表论文
1. 福井拓（Taku Fukui）、宗方聪（Satoshi Munakata），"通过从包含图像的文档中生成检索用洞察以提升多模态检索增强生成系统的检索精度"，收录于日本人工智能学会第39届年会（JSAI2025）论文集，4Q1-GS-10-01，2025年。
2. 鸣田贤一郎（Kenichirou Narita）、宗方聪（Satoshi Munakata），"Chunk-Link：基于上下文感知的分块补全"，收录于第2届知识图谱检索增强生成研讨会（RAGE-KG 2025）论文集（CEUR Workshop Proceedings，Vol-4079），short1，2025年。

## 授权协议
CC-BY-4.0

## 联系方式
鸣田贤一郎（Kenichirou Narita）：k.narita{at}fujitsu.com