Day 59 of #100daysofnetworks
Graph Memory: Complete, Contextual, and Time-Aware
Hello everyone!
We have made more progress on our Knowledge Graph since the previous article, and I am excited to show this and discuss what it means and where it will lead us next! Lately, every article is a continuation of the previous article, and we have been doing a lot of work with Knowledge Graphs and GraphRAG for Artificial Intelligence.
In the previous article, we had some impressive breakthroughs.
We were able to make our knowledge graph 6x bigger, 100% lower cost, and 540x faster. In plain English:
We used 12,000 articles from the Artificial Life dataset instead of 2,000.
It cost us $0 instead of $20.
It took less than one minute instead of five hundred and forty minutes.
And it set us up very well for iteration and improvement!
Since then, I have taken the Knowledge Graph even further!
Latest Improvements and Reasoning
This week, I have made two improvements to the Artificial Life knowledge graph. I have kept the Cognee one as a secondary, so that I can compare mine against it in the future. Here are the improvements to my latest approach:
I have added temporal elements to the knowledge graph:
Year nodes
date_published node properties
I have added to the completeness of the Paper nodes, adding more fields from the original dataset. These are the current properties/attributes for Papers:
date_published
title
category
authors
url
summary
With these two improvements, our Knowledge Graph has now become a Temporal Knowledge Graph. It has been given memory capabilities. Instead of having 12,000 Artificial Life articles and no temporal awareness, I can do things like:
Create ‘living system’ AI interfaces that gradually forget old memories, behaving more like life.
Analyze how ecosystems change over time.
Predict what will happen in the future based on what has happened in the past.
And much, much, much more. Creativity and imagination is the limitation.
Let’s see what we can do, now!
Show and Tell Demonstration
First of all, exploring the Knowledge Graph has become much easier, in Neo4j. There is an “Explore” mode which is very useful, and now it is even more useful, because I can start by exploring individual years.
If I zoom in, there are a lot of Year nodes, and I can see individual years.
I can choose one, dismiss all of the others, and then expand it. Let’s see what research papers were written in 2009.
Let’s look at one node and its properties!
Super clean! All of the fields landed well. Let’s expand this node to make sure the People nodes came through.
Perfect! Let’s try some queries!
Excellent! I can put this data together in a familiar format, rows and columns! Look at that scrollbar! I can look through Artificial Life papers to my heart’s content! I can see the oldest papers from 1993, and I can see the latest papers that were identified when I created this dataset.
I can see the collaboration networks that happened, by year.
That is what it looks like when a lot of researchers work together on very few papers.
I can also see the author-paper networks by year, which is another perspective.
There’s a lot of activity! If you look at those two clusters, you can see that they are clusters of people working on two papers.
You can see that happening here. Lots of people, and one paper.
Success! What’s Next?!
It worked! Our temporal expansion worked. Next, I will be adding Category context, and then some AI enrichments. This is probably already good enough for GraphRAG uses, but I am not done, yet! That’s all for today!
Please Support this Work!
I have written over 50 articles for this series. Each one takes about four hours of research, and several pages of writing and editing. Here are some ways you can support the blog!
Please subscribe if you have not. This motivates me like nothing else!
LET’S DO BUSINESS. Reach out to me if you need data or AI help! Happy to help! You can read about the partners we are currently looking for here.
BIGGEST HELP to the BLOG: Please consider upgrading if you are a subscriber. Thank you to all current paying subscribers for making this research and development possible!
Please buy my book to understand how I think about Natural Language Processing and Network Science combined.
Please reach out to me if interested in training related to Data Science, Data Engineering, Network Science, Knowledge Graphs, Artificial Intelligence, or anything else I write about. Feel free to message me on Substack or on LinkedIn!
Feel free to hang out in the comments and have a good time!
We have come so far since the very first day of the very first #100daysofnetworks. I love writing for this series. Thank you for being a part of it!