Gemini turns knowledge-graph history into GraphML
Idiaye reports researching knowledge-graph history with Gemini Deep Research and converting the report into an interactive graph and GraphML through a publicly shared notebook.
linkedin.com
Original prompt
Shared by the source author. See source notes for context.
Research the history of knowledge graphs
Reach people exploring AI agents
See placements →Evidence at a glance
Sourced account- Source author
- Ifeanyi Idiaye
- Original account
- Read the source ↗
- Published
- Source link checked
- Reported result
- Idiaye reports researching knowledge-graph history with Gemini Deep Research and converting the report into an interactive graph and GraphML through a publicly shared notebook.
- Human involvement
- Idiaye publishes the notebook and output formats. A public reply criticises broken references; the report citations and extracted relationships have not been independently validated here.
A linked account supports attribution. It does not mean we independently reproduced its result.
Report an issue with this case
Open correction emailOpens your email app with the case link included. Nothing is sent automatically.
How the agent did it
- Start a background Gemini Deep Research interaction with web search.
- Poll for completion and retain the resulting report text.
- Use KGGen to extract entities and relationships from that report.
Ifeanyi Idiaye shares a Gemini Deep Research experiment about the history of knowledge graphs, together with the executed notebook used to transform its output. The notebook starts a background research interaction with a search tool, polls for completion and extracts the final report text.
It then passes the report to KGGen using Gemini to extract entities and relationships. NetworkX supports the graph layout; the notebook produces an HTML visualisation and writes a GraphML file for further graph work. The concrete result is thus a research report plus inspectable graph artifacts, not just a picture of a proposed architecture.
The extraction stage inherits the report's factual limitations. A public reply criticises its reference links as broken, and this review has not independently validated all report citations or graph edges. An attractive network can therefore organise claims without proving them. The copyable workflow keeps the intermediate report available, exports relationships in a standard format and treats citation and edge checking as separate work. The original prompt is a short research instruction; the public notebook supplies the implementation detail that the prompt alone does not contain.
Prompt
Original prompt
Source: LinkedIn · Ifeanyi Idiaye
Frequently asked questions
What files does the notebook produce?
It renders an HTML graph and exports GraphML after extracting entities and relationships from the research report.
Does an exported graph prove the facts?
No. Extraction depends on the report, and graph edges require their own source checks.
What citation issue was raised?
A public reply says the references are broken. This account does not independently resolve every linked citation.
More details
Remaining steps
- Render an HTML graph and export GraphML for inspection.
- Check source links and graph relationships before relying on them.
About this account
Gemini performs a background research investigation; a subsequent KGGen/model stage transforms the completed report into graph data and exported artifacts.
Firsthand account by Ifeanyi Idiaye; original publication date verified against Public LinkedIn JSON-LD datePublished 2026-01-05T17:45:04.556Z; public executed KG_Research_Agent.ipynb inspected. Outcomes are the author’s reports, not independently reproduced. Original research prompt shared in the post. The public notebook at github.com/Ifeanyi55/GephiDatasets confirms the research, extraction and export steps.
Credit: Ifeanyi Idiaye
Source published
Added to Agents In Use
Source checked
Tools and topics
Gemini Deep Research API · KGGen · NetworkX · HTML visualisation · GraphML · Google Colab
Knowledge-graph history · Graph extraction · Citation checking
Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.