Skip to content

Home

Public Knowledge Graph
Public datasets of "Usable Knowledge" generated using Fujitsu's Knowledge Graph Enhanced RAG (Retrieval-Augmented Generation)

Challenges in using data with generative AI

(画像に書かれているテキスト)

When searching and analyzing internal/external data and websites with generative AI, much of the target data is "unstructured data." As a result, outputs are often based only on fragmented knowledge, and hallucinations can occur. Therefore, it is important to structure data and turn it into knowledge, but there have been very few such initiatives.

Significance of publishing knowledge

(画像に書かれているテキスト)

We structure highly beneficial public data (manuals, documents, datasets, etc.) into a knowledge graph using Fujitsu Knowledge Graph Enhanced RAG technology, and publish that knowledge. By doing so, we hope people will recognize the value of structured knowledge and work together with us to solve challenges in data utilization.

Value provided by the Public Knowledge Graph

  1. Use of public data
    • By structuring highly beneficial public data, we make it easier for anyone to utilize data.
  2. Contribution to research activities
    • Researchers can use it as benchmark data to support research and development of AI technologies.

Technology overview

Target industries and users

  • General users and AI researchers who search and analyze internal/external data and websites using generative AI

Challenges in target industries and operations

  • When searching and analyzing internal/external data and websites with generative AI, much of the target data is "unstructured data". As a result, outputs are often based only on fragmented knowledge, and hallucinations can occur.

Issues with existing initiatives

  • Until now, data has been structured in proprietary formats or in Resource Description Framework (RDF) format for each specific purpose (e.g., Google Data Commons, Discourse Graphs, published knowledge graphs). However, these efforts have mainly targeted analytical data handled by a limited number of researchers, and cannot be regarded as broad initiatives to structure general unstructured data used in everyday contexts.
  • Fujitsu's differentiating technology, Fujitsu Knowledge Graph Enhanced RAG, can automatically create knowledge graphs as structured data. Up to now, however, we have applied this technology mainly to highly confidential data (Fujitsu internal data and customer data), and have not targeted public data.
  • Thus, existing initiatives alone have not been sufficient to convert highly beneficial public data (manuals, documents, datasets, etc.) into a form that can be used as knowledge.

Solution

  • To fundamentally solve the challenges of data utilization, it is important to structure highly beneficial public data and convert it into knowledge that anyone can easily handle. We have therefore started an initiative to convert public data into knowledge graphs using Fujitsu Knowledge Graph Enhanced RAG and to widely publish that knowledge.

(画像に書かれているテキスト)

Fujitsu's technological advantage

  • For details on the advantages of Fujitsu's "Knowledge Graph Enhanced RAG" technology used to create and utilize this knowledge, please refer to the technical blog posts listed under "Related Information".

Try the technology