---
title: "Knowledge Graph Notes: How to Connect Ideas Across Sources"
description: "Knowledge graph notes link ideas across articles, papers, videos and podcasts. Learn what to capture, how to link it, and how to keep the graph useful."
canonical: "https://omphalis.ai/blog/knowledge-graph-notes"
source: "https://omphalis.ai/blog/knowledge-graph-notes"
---

# Knowledge Graph Notes: How to Connect Ideas Across Sources

Updated November 16, 2026 · 15 min read

![Knowledge Graph Notes: How to Connect Ideas Across Sources](https://media.babylovegrowth.ai/blog-images/organization-18678/1794647240182_Isometric-note-cards-linked-across-geometric-planes.jpeg)

Knowledge graph notes are notes where each idea sits in its own small note and is linked to the other notes it relates to, so your reading forms a network instead of a pile. The part that makes the network useful is not the picture of dots and lines. It is the sentence you write on each link that says why two ideas belong together, and the trail back to the exact passage in the source. We built Omphalis around that second part: you read, annotate and ask questions inside the original text, so every note stays attached to the passage that produced it.

## TL;DR

- A knowledge graph for notes is small notes (nodes) plus the relationships between them (links). A personal one is simply the graph you build from your own reading.
- Start with one question, not a tool. Capture source notes with locations, then write idea notes in your own words and link them with a sentence of explanation.
- Typed or explained links beat bare links, because they tell you why two notes are connected without reopening both.
- Add only links you would actually follow later. A graph where everything touches everything is no better than a folder.
- Keep the source one click away. Notes that reopen the original passage are the ones you can trust months later.

## Table of Contents

- What knowledge graph notes actually are
- Why a graph beats folders for lengthy reading
- Build your own knowledge graph from articles, papers and videos
- What to capture from different kinds of sources
- Keep the graph useful, not just large
- Ask questions across your saved sources
- Where Omphalis fits
- Our perspective
- FAQ
- Sources

## What knowledge graph notes actually are

A graph, in the mathematical sense, is a set of points joined by lines. A [2022 North Carolina State University Linux Users Group presentation](https://lug.ncsu.edu/presentations/2022-02-25-note-taking-with-graph-theory.pdf) describes these as vertices and edges, and points out that the old Zettelkasten slip-box method, with numbered notecards that reference one another, already works like a directed graph: each card is a vertex and each reference is an edge that points one way.

So "knowledge graph notes" is not a new invention. It is the slip-box idea applied to whatever you read now: web articles, PDFs, lecture videos and podcasts.

Three terms get mixed up, so here is the plain version:

- **Linked notes**: notes that point to each other. The link says "these are related" and nothing more.
- **Knowledge graph notes**: linked notes where the relationships carry meaning, such as "supports," "contradicts" or "is an example of," and where each note holds one idea.
- **Personal knowledge graph**: the graph that belongs to one person. Researchers who write about this define it as the entities and relationships that matter to one researcher, from the papers they read to the methods, tools and datasets those papers use, as described in the academic paper on [personal research knowledge graphs](https://arxiv.org/pdf/2204.11428).

If you have seen IBM-style diagrams of enterprise knowledge graphs, those are the same structure at company scale: entities and the relationships between them. You do not need a database or special software to use the principle. You need a habit of writing one idea at a time and saying how it connects.

## Why a graph beats folders for lengthy reading

Folders force a note into one place. A paper on sleep and memory could belong under "psychology," "study skills" or "chapter 4 reading." You pick one and the other two doors close.

The NC State guide makes the case for organizing by concept rather than by date or course. It recommends notes that are atomic (one idea), concept-oriented and densely linked, and it prefers associative structures over rigid hierarchies, because connecting notes is what reinforces what you know. That matters most for the people who struggle with long reading:

- **Students** collect dozens of papers per course and need to see how one author's claim answers another's.
- **Professionals** curating industry insight need to find the three sources that support a recommendation, fast.
- **Podcast listeners** want to return to one moment from a two-hour episode and see what else in their notes it touches.
- **Anyone overwhelmed by long texts** needs a way to keep the useful part without rereading the whole thing.

In each case the value is in retrieval and connection later, not in capture today.

## Build your own knowledge graph from articles, papers and videos

The workflow below follows the same shape as a published [guide to knowledge graph note taking](https://fenbs.ai/blog/knowledge-graph-note-taking): pick a question, preserve sources, write ideas in your own words, link them with reasons, then summarize. We have added where it goes wrong and where Omphalis takes work off your hands.

![Five steps for building knowledge graph notes](https://media.babylovegrowth.ai/blog-images/organization-18678/1794647286888_Five-steps-for-building-knowledge-graph-notes.jpeg)

### Step 1: Choose one question

Pick something you are actually trying to answer: "Does spaced practice help adult learners?" or "What do the sources say about remote work and productivity?" A graph without a question grows in every direction and helps in none.

Where it goes wrong: starting with a topic like "psychology." A topic has no edges. A question does, because every source either supports it, challenges it or adds context.

### Step 2: Save the source and keep its location

For each article, paper, video or episode, make a source note: title, author, date, where it lives, and the passages you used with their locations (page, section, timestamp). Attribution matters. Specifications for structured note vaults, such as the [Mosaix Ingest format](https://mosaixformat.org/skill/mosaix-ingest/SKILL.md), require that every note record where its content came from so the source stays traceable.

Where it goes wrong: copying a quote without its location, then spending twenty minutes hunting for it.

This is the step Omphalis handles most directly. You can import PDFs, EPUBs, Word documents and feeds, and save articles, papers, podcasts and videos into one library. Content is cleaned and structured for reading, and your highlights and notes are attached to the original passage, so a note [reopens the passage](https://omphalis.ai/highlights) it came from. If you want the detail on tying notes to quotes, we wrote up [how to attach notes to quotes](https://omphalis.ai/blog/attach-notes-to-quotes) so you can find them later.

### Step 3: Write idea notes in your own words

Now write notes that each hold one claim or one question. Good guides recommend atomic notes, a record of why you captured something, and a reference back to the original context, then refining the note when you revisit it. A raw capture typically yields a handful of insights, not dozens. One published enrichment workflow describes [3 to 8 atomic insights per source](https://www.dsebastien.net/the-knowledge-ingestion-process/), together with metadata such as author, date and source URL.

An example from a student's reading on memory:

- Source note: a review article on retrieval practice, section 3.
- Idea note: "Testing yourself beats rereading for long-term recall."
- Why I captured it: it answers my question about exam prep.

Where it goes wrong: writing a note that is a paragraph of the author's wording. If you cannot say it in your own sentence, you have not understood it yet. This is where reading support matters. When a passage is dense, Omphalis explains it inline so you can [understand a hard paragraph](https://omphalis.ai/reading-comprehension) without leaving the document, and you then write your note once the idea is clear. Our approach is designed to keep you with the full source rather than replace it with a shortcut, because a note built on a misread passage links everything to a mistake.

### Step 4: Link ideas, and write why

This is the step that turns notes into a graph. For each idea note, link it to the source that supports it, and to any note that challenges it, and add a sentence on the relationship.

One 2026 post argues that [typed edges](https://www.mejba.me/blog/infinite-brain-knowledge-graph-ai) such as supports, contradicts, derived from, part of and authored by are the real difference between a graph and ordinary wikilinks. Its reasoning: an untyped link makes you read both notes in full to learn why they are connected, while a typed one lets you follow only the relationships relevant to your question.

You do not need ten relationship types. Three cover most research reading:

| Relationship | Plain meaning | Example |
| --- | --- | --- |
| Supports | This source gives evidence for the idea | Review article supports "testing beats rereading" |
| Challenges | This source complicates or contradicts it | A study on very short intervals challenges it |
| Applies | This idea shows up in a specific setting | Retrieval practice applied to medical exam prep |

Where it goes wrong: linking because two notes share a keyword. Ask whether the link answers "why are these connected?" If you cannot write the sentence, skip the link.

Across sources, you can also let the software point out candidates. Omphalis connects related content in your library, and our guide on how to [evaluate suggested connections](https://docs.omphalis.ai/learn/intermediate/suggested-connections) shows how to accept the ones that matter and ignore the rest. You stay the judge of each link.

### Step 5: Write an overview note

Finish with one note per question: the question, your strongest conclusions so far, the gaps, and the next thing to investigate. This is the page you open first. It also solves the "blank page" problem when it is time to write the paper, memo or presentation.

## What to capture from different kinds of sources

The structure is the same across formats, but the capture details differ.

**Articles and papers.** Record the claim, the evidence type (experiment, survey, opinion) and the author's stated limits. For research, the entities worth tracking are papers, datasets, concepts, methods and people, and relationships like cites, applies, supports or contradicts, along with attributes such as publication year and method, as one [knowledge ingestion guide](https://www.dsebastien.net/the-knowledge-ingestion-process/) describes. Our walkthrough on how to [take notes from articles](https://omphalis.ai/blog/take-notes-from-articles) covers this in more detail, and our [research notebook templates](https://omphalis.ai/blog/notebook-templates-for-research) give you a ready layout.

**Videos.** Capture the timestamp with the idea. A note that says "at 14:20 she explains why the sample was biased" is worth ten that say "good video." We describe a [hybrid workflow for annotating YouTube videos](https://omphalis.ai/blog/annotate-youtube-videos) that ends in study-ready notes.

**Podcasts.** Spoken content is the easiest to lose. Mark the moment, write the idea in a sentence, and link it to related notes the same day. Our [podcast note-taking system](https://omphalis.ai/blog/podcast-note-taking) shows how to do it without pausing every thirty seconds.

**Your own writing.** Your drafts are part of the graph too. Notes about how other people's ideas relate to yours are the most valuable ones, and we explain how to keep [your own writing and other people's reading](https://omphalis.ai/blog/keeping-your-own-writing-and-other-peoples-reading-in-one-place) in one place.

A practical way to organize all of this into reusable notes is covered in a roundup of [note taking methods for students](https://lizardslunch.com/note-taking-methods), which compares several systems you can borrow structure from. For researchers and writers, a [source verification workflow](https://hellenismwatch.org/en/blog/source-verification-methods) is a useful companion habit: it shows how to trace each claim to its origin so that every link in your graph rests on something you checked.

![What to capture from different kinds of sources](https://media.babylovegrowth.ai/blog-images/organization-18678/1794647325758_What-to-capture-from-different-kinds-of-sources.jpeg)

## Keep the graph useful, not just large

A graph can fail in two opposite ways, and both are common.

**Too loose.** Free-form notes are quick but hard to query. A published appendix on personal knowledge graphs describes a spectrum from free-form note taking (no rules, fast, little meaning) to structured knowledge (more rules, slower, but consistent and easy to validate and query), and cites the idea that [a little semantics goes a long way](https://dmccreary.github.io/ai-for-ee-research/appendices/pkg/). In practice: a few consistent link types are enough.

**Too dense.** If every note links to every other, the graph stops telling you anything. The same 2026 post on typed edges notes that over-connection can make a graph uniformly dense and no better than unstructured notes, and suggests a test: [would a future question actually follow this link](https://www.mejba.me/blog/infinite-brain-knowledge-graph-ai)? If not, leave it out.

Two more habits keep a graph healthy:

1. **Distill quickly.** A post summarizing Andy Matuschak's practice reports that a capture not distilled within 48 hours usually never will be, and recommends mostly writing notes about your own ideas and how others' ideas relate to them, as covered in [this write-up](https://www.fzeba.com/posts/64-graphs-for-learning/). Keep your review loop tight.
2. **Pair structure with recall.** The same write-up distinguishes a structure loop (capture, atomize, link, map) from a retention loop (prompt, schedule, recall, grade). A beautiful graph you cannot recall from is a filing system, so revisit your notes on purpose. Our piece on the [digital commonplace book](https://omphalis.ai/blog/digital-commonplace-book) shows how revisiting notes builds retention, and a weekly slot of [5 to 15 minutes](https://omphalis.ai/blog/organize-notebooks) is enough to keep notebooks tidy.

## Ask questions across your saved sources

Once you have a few dozen sources, the best test of your graph is a question: "What do my sources say about X, and where?" In Omphalis you can ask questions about your content and get answers with citations from the source, so you can jump to the passage that supports each claim instead of trusting a summary. If you are weighing what to look for in a tool that does this, see our guide to an [AI research tool for questions across saved sources](https://omphalis.ai/blog/ai-research-tool-for-questions-across-saved-sources-with-citations). To find new material to feed the graph, our docs show how to [search for sources on a topic](https://docs.omphalis.ai/learn/intermediate/discover-anything).

## Where Omphalis fits

Omphalis is a platform for saving, reading, annotating and engaging with articles, papers, podcasts and videos. It cleans and structures content, explains complex passages, attaches your notes to the original text, connects related content, lets you listen with natural voice narration, and exports your work and organizes it into notebooks. The Free plan lets you start without a published price. Pro is $9 per month or $90 per year, Premium is $19 per month or $190 per year, and Scholar is $39 per month or $390 per year. Unit packs start from $10 one-off. Paid plans add features such as deep reading, PDF extraction, advanced listening and import and export. See the [pricing page](https://omphalis.ai/pricing) for what each plan includes.

![Omphalis](https://media.babylovegrowth.ai/blog-images/organization-18678/1790615591899_omphalis.jpg)

If your reading is already scattered across tabs, PDFs and podcast apps, begin with one question and import the first five sources. Our page on the [commonplace book app that keeps the source](https://omphalis.ai/commonplace-book) shows the same idea in a single notebook.

## Our perspective

> Most note systems fail at the same moment: when you return to a note and cannot tell what it was about or where it came from. A graph makes that worse if the links are decoration. We think the source is the anchor of the whole structure. That is why Omphalis keeps you inside the original text: explanations appear next to the difficult passage, your note sits on the sentence that prompted it, and questions come back with citations. The aim is understanding, not speed. A tool that hands you a summary may feel fluent, but a graph built on summaries links ideas you have not yet understood. So build your graph slowly. Link only what you can explain, keep the passage one click away, and revisit it on a schedule. Done that way, connections across sources stop being a filing project and become how you think. *Omphalis Team*

## FAQ

### What is the difference between a knowledge graph and linked notes?

Linked notes only say that two notes are related. A knowledge graph adds meaning to each link, such as supports or contradicts, and keeps each note to one idea. That extra sentence of explanation is what lets you find the right connection months later.

### Do I need special software to build a personal knowledge graph?

No. The method works with any tool that lets you write small notes and link them, and the Zettelkasten method did it with paper cards. What helps is keeping each note tied to its source passage, which is what Omphalis does with annotations attached to the original text.

### How many links should each note have?

As many as you can justify, usually a few. The test is whether a future question would follow the link. If you cannot write a sentence on why two notes are connected, the link adds clutter.

### How do I connect notes from videos and podcasts to notes from papers?

Capture the timestamp and the idea in one sentence, then link it to the paper note it supports or challenges, with the reason. Treat the episode as a source note like any article. Importing the content into one library makes the links easier to find.

### How long does this take?

Capture is quick if you write one idea per note. The habit that matters is a short weekly review, and 5 to 15 minutes is enough to keep a notebook organized.

## Sources

- [Note-taking with the power of graph theory (NC State Linux Users Group, 2022)](https://lug.ncsu.edu/presentations/2022-02-25-note-taking-with-graph-theory.pdf)
- [Personal Research Knowledge Graphs (arXiv, 2022)](https://arxiv.org/pdf/2204.11428)
- [Knowledge graph note taking: connect useful ideas](https://fenbs.ai/blog/knowledge-graph-note-taking)
- [Mosaix Ingest specification](https://mosaixformat.org/skill/mosaix-ingest/SKILL.md)
- [The Knowledge Funnel: From Capture to Connected Notes](https://www.dsebastien.net/the-knowledge-ingestion-process/)
- [Infinite Brain Knowledge Graph vs PARA for AI Retrieval](https://www.mejba.me/blog/infinite-brain-knowledge-graph-ai)
- [Personal Knowledge Graphs, AI for EE Research](https://dmccreary.github.io/ai-for-ee-research/appendices/pkg/)
- [The Graph Is Not the Point: Building a Personal Knowledge Graph](https://www.fzeba.com/posts/64-graphs-for-learning/)
- [Note Taking Methods For Students: 10 Systems To Try](https://lizardslunch.com/note-taking-methods)
- [Replicable Source Verification Methods for Researchers and Journalists](https://hellenismwatch.org/en/blog/source-verification-methods)

## Recommended

- [Apps that connect ideas across your saved articles (2026)](https://omphalis.ai/blog/apps-that-connect-ideas-across-your-saved-articles-2026)
- [Compare What Research Sources Say About One Question](https://omphalis.ai/synthesize-sources)
- [30–60 Day Action Plan to Build a Second Brain for Knowledge Work](https://omphalis.ai/blog/build-a-second-brain)
- [How to Attach Notes to Quotes So You Can Find Them Later](https://omphalis.ai/blog/attach-notes-to-quotes)

---

Listed on: [Featured on launched.tools REVIEWED ✓](https://launched.tools/tools/omphalis)
