'Semantic Tree' -- What Did He Mean?

“It is important to view knowledge as sort of a semantic tree — make sure you understand the fundamental principles, ie the trunk and big branches, before you get into the leaves/details or there is nothing for them to hang on to.”

~ Elon Musk

This quote packs a fundamental insight into learning and how our minds work that is rather beautiful. Unfortunately, I think for many it’s not immediately obvious what it means. I did some digging, but wasn’t particularly thrilled about the descriptions on the internet. So, I thought I would give unpacking this statement a go. This is my understanding of Elon Musk’s semantic tree – hopefully it will be of help to others.

Of Indices and TOCs 🐝

First let’s review some ordering mechanisms that we are familiar with: Indexes and tables of contents.

Indexes are an alphabetical ordering of words that point to page numbers or locations of appearance of the word in question. Tables of contents are a topical ordering1 of some body of work by its major parts. These are two primary navigational methods that allow us to explore works, especially books, quickly.

In some sense, Musk is saying to create table of contents of the things you are learning.

From Word Salad To Semantic Tree 🐝

Back to those trees.

Trees are a type of structure that appear in many many places, especially in natural language and computer science. In natural language, “language arts”, or English class, many will remember “diagramming sentences”. When you diagram a sentence, you are parsing the function of each word and mapping it to a tree that illustrates the structure underlying those functions.

But trees can be used for more than just grammar lessons. In computer science, trees are a fundamental type of data structure that appear everywhere. Say we want to organize some words for that index we mentioned earlier. Typically, we’ll extract key terms and sort them alphabetically. You can visualize this re-ordering as a tree structure2.

The way this tree works is that a word is built character by character as you drill down deeper into the tree. Each time you drill down a level you select the next character in the word from all possible choices – in our case a-z. Each node, or location in the tree, represents a word and each word’s letters are constructed from the links that must be passed on the way to that word’s node from the root.

For example, to build the word “fox” the tree path might look like…

(root) --f--> (f) --o--> (fo) --x--> (fox)

…and “dog” would be built from…

(root) --d--> (d) --o--> (do) --g--> (dog)

Given that our nodes at each level are ordered alphabetically from left to right, words built from this tree will also be successfully ordered alphabetically from left to right. You might say, when we sort things alphabetically, we are building this tree structure implicitly in our mind to create the sorted list.

Sorting words by letter makes words easier to search, lookup, and reference. But it doesn’t really help us understand the material, the meaning behind, or relationship between these words. This is where semantics comes in: Say instead of using a tree to order words alphabetically, we sorted words based on their meaning; their semantics. Instead of selecting the next node in the drilling process based on the next character, we select nodes based on inherent attributes or properties that are related to the current node.3

Textbooks are great examples of these semantic trees, since they are topically ordered: Well-designed textbooks are often sensibly arranged based on the materials’ inherent properties and highlight keywords important for the given section. You could build a topical tree of terms out of the category the textbook belongs to, its over-arching topic, then its chapters, sections, and lastly its keywords.

By going through the process of building these trees yourself, you are explicitly defining the relationships between all of the interrelated concepts (nodes in the graph). You are building a better understanding of those concepts, the relationships between them, and the space they inhabit. In turn, it seems to me understanding of the concepts will be more thorough, robust, precise, and easier to recall.

Make Like A Tree and Leaf 🐝

The description above outlines “the trunk and branches”. Once you have a semantic tree built from keywords, terms, and concepts as described above, you’ll find that it’s easier to remember related details and minutae.4

The leaves simply hang on the tree.

Mind Growth 🐝

We hairless apes, with our spongy cortexes, evolved to use language. By taking the words and terms of said language and deliberately sorting them into coherent semantic trees, we can better [[untangle]] and understand the meanings behind the words. And the less tangled, more coherent our understanding of words, topics, ideas, and concepts, the farther and greater we may grow. The sturdier the trunk and the wider the breath of the branches, the more leaves can thrive.

And with more thriving ideas, the more likely we may bear fruit.

  1. Mentioned in how to read a book

  2. Ok, so using a tree structure to alphabatize a list is a bit contrived in terms of what data structures we could use, but the point is to illustrate how trees can order data to those without a computer science background. 

  3. Diagramming a sentence like “The fox jumped over the lazy dog” yields a semantic tree very similar to a grammar tree since the words are conceptually so simple. But it’s important to keep in mind that there are several ways to determine specific placements and there will doubtless be disagreement about precise details, especially for more complex words and ideas. 

  4. Noteworthy: As a data structure, trees are faster to traverse than non-tree-graphs.