What is the meaning of a word? There’s always talk about how computers lack to ability to associate meaning with a word, but is meaning really being defined anywhere? I read an article about new software tools that create parse trees of essays and compare them with previous parse trees in an attempt to produce a meaningful summary. Is that what we as humans do? If not, what do we do?

Let’s look at this example: Tonight I was looking at the salary field of a job posting, and it said “DOE”. Having never seen this before I immediately thought “dead on arrival,” but that’s another essay in and of itself. A moment later though, the phrase “depends on experience” popped into my head. Now here we go…

Let’s look at what happens in my brain as I ponder.

DOE

First and foremost, I have to recognize this is an acronym, not just the word “doe.” Once I do, I know I’ve seen positions that list salary as being based on experience. I know that salaries are in fact based on experience, and therefore that’s the first thing that jumps into my head.

DO Experience

Next the O immediately becomes “on”. I know the fragment “on experience” from other adds and such, so I assume that the salary is “something on experience.”

D on Experience

Finally, the phrase “depends on experience” is flashes across my mind, and I conclude that this is the most probable expansion of the acronym. There could be subtle differences, such as “dependent on experience”, but in the end it doesn’t matter which one it really is because the meaning has been ascertained.

Although this seemingly trivial process takes all but two seconds, in reality it is extraordinarily complex. In order to understand the relations between salary and experience, we first need to understand the meaning of salary and the meaning of experience. And it is that understanding that is so elusive, because we lack a proper definition for meaning in this context. In order to better understand what meaning is, we’ll compare it to a common spoken word example during a verbal communication.

“I think this is the end of the road for the Eagles this season.”
“What do you mean?”
“I mean, I don’t think we can even make it to the playoffs.”

In the example above the second participant asks the first to clarify his statement. But what his statement actually is, is an invitation for inference. We can think of his first statement as a representation of his third. He is not actually saying what he means, he is using a representative to express it. This same idea can be applied to any words, ideas, memories, etc., which are abstract representations of actual things. (This is an extremely simplified statement. Assume that “actual things” can be either fictional or non-fictional. For instance, “unicorn” represents something that is both real and fictional.)

It becomes clear that no matter how much parsing a computer does, or how good its guessing becomes, type in the word “tree” and it simply doesn’t associate it with an actual tree. Part of the problem is the fact that we as humans have five senses that compare and contrast our “real world” in parallel. For instance, it’s one thing to describe a tree to a person who has no idea about a tree. It’s another to have them see it, touch it, smell it, and so on. In order to really create “meaning”, the person needs to gather data through all five senses and associate the word with the object.

Until a computer can truly associate one thing, all of the data you can possibly pump into it will be just that, data. You can run sophisticated algorithms on it, get word counts, all kinds of crazy analysis that would take a human an eternity to do. But it will never know that it doesn’t know anything.