This page was translated from French to English with the help of AI. Read the original in French.
Here’s the conclusion of my talk at Que Du Web 2017 in Deauville.
I fast-forward through what we covered in the previous weeks, and we draw the lessons from it.
(See the overview of the talk with all the episodes)
The matching slides:
The transcript (heavily summarized and simplified)
Conclusion:
We didn’t invent anything.
A lot of methods that look modern are actually as old as the hills.
Information and knowledge seem to organize themselves naturally into trees and graphs.
It’s a representation that makes the complexity of certain phenomena easier to grasp.
And we shouldn’t always try to boil that complexity down to an average, a number, a table of numbers, some stats.
It may feel more reassuring, but it’s wrong.
Look at the TextRank example, where the graph captures more relations between words than purely statistical models do.
Who knows what this picture shows?
The year is 1859, and Darwin publishes “On the Origin of Species”.
A 300-page book that he wants to reach the general public, not just specialists.
And he insists on including a graph in it, his “tree of life”, of which you can see a sketch here.
And it’s the ONLY illustration in the entire book.
A representation so telling that he can’t express his idea any other way.
Today we know that not only are graphs in tune with the way our brain works (relations, analogies, links, symbols),
but they’re also, as we saw with Leonhard Euler, a powerful mathematical tool that lets us do things other branches of mathematics couldn’t.
So we can apply them to concrete problems, like analyzing texts or website structure.
Beyond an “artistic” visual representation, these graphs reflect essential properties of the objects they model.
That’s it. I hope I’ve intrigued you enough to make you want to dig deeper, experiment, and ask lots of questions!
Thank you!