What can computer image processing teach us about language? (Part one)
Last night, I was laying in bed, idly thinking about a project for my family’s printing business. The problem is as follows:
In order for us to replicate an image on some other medium, we need a copy of the original image. Inevitably, our less technologically saavy customers will send us tiny, highly compressed files, later asking for a large print with that same image. This is equivalent to hiring a mural painter to do your living room wall, and handing them a postage stamp to work from. File compression is a big problem in the printing business, and something that few non-graphics people really understand.
At this point, I sat up in bed and was struck that this presents a beautiful metaphor for language, on several different levels. The following is an exploration of this metaphor. For the purposes of this article, I snapped a picture of a few, not-so-randomly selected good books, and I’ll be using the same picture throughout the article. The books, in case you’re wondering, are Language, Thought and Reality by Benjamin Whorf, The Brothers Karamazov by Fyodor Dostoevsky, and Far from the Madding Gerund, the Language Log book.
The Original Image
So let’s start from a flawless original. In order to show you the original, I’ve uploaded it as a TIFF file. Now, the beauty of the TIFF file format is that absolutely nothing is lost when you save the picture. Every bit of detail that was there in the camera is there in the TIFF, earning TIFF the designation of being a “lossless” format. However, this is not without a price. Let’s view the file now:
Because this is a lossless file, it is a very large and cumbersome. For this small picture, the TIFF file was a little more than 1.8 Megabytes. This is why it probably took some time to load for you. In addition, they’re difficult for some software to open. Firefox has to use Apple’s Quicktime to open them, and it can take a while even for a powerful program like Adobe’s Photoshop to open them. So, there’s a compromise here. You can get a high fidelity, lossless file, but it takes a long time to transmit, load, and open it.
Quality over Cost
The next option (that I’m going to explore) is the JPEG file format. JPEG is a “lossy” format, meaning that, in order to save space and time when opening it, it throws out some of the data and detail. Ideally, it’s only throwing out details that are unnecessary or invisible to the human eye, but often, some loss becomes quickly apparent. Below is the highest quality JPEG that Photoshop can make:
This file (compressed at Maximum quality) is pretty detailed still, and little is visibly lost. However, the filesize has gone down from 1.8MB to 188kb. This is a very good compromise, because you can still get the detail across, but you don’t have to take up as much time and space to work with the file. Now, let’s check out the same file, compressed at 50% quality:
Here, you start to see some “artifacts”, or little jagged patches in the solid colors (look around the text on “language, thought and reality”). However, from this loss of quality, we are able to shrink the file down to 40kb. Finally, just for grins, let’s look at 0% quality JPEG (the most compression with the least quality):
We’ve now hit a very, very noticeable loss in quality, although the file is only 16kb in this state.
Here too, we see evidence of the compromise between size and detail.
What does this have to do with language?
Compare the following four example interactions:
Kim: “Where’s Mom?”
Pat: “Although I’m not entirely sure, as I’ve been out of contact with her, I have no means of locating her exactly (via GPS or otherwise), and she has been known to make stops unannounced, she had earlier expressed an intent and desire to go to the John’s Hair Salon, on 28th Street. Considering that she left around 10am, it is now 10:15am, and her salon sessions usually last approximately one hour, there is a good chance that she is still currently at the Salon.”
Kim: “Where’s Mom?”
Pat: “She said she was going to John’s Hair Salon when she left 15 minutes ago.”
Kim: “Where’s Mom?” Pat: “At the salon.”
Kim: “Where’s Mom?” Pat: “Out.”
Here, we have a very similar situation to the one above. There is a block of information that needs to be expressed, and many options as to how to best express it to maximize detail and efficiency.
Answer one is the Maximum Quality JPEG file of verbal expression. It gives every bit of necessary detail (and more) at the expense of time and energy. However, Pat’s response violates both Grice’s Maxims of Quantity and Manner, giving FAR more information than required or desired, slowing down interaction and cluttering Kim’s mind with more detail than needed. If everybody talked like this, nothing would ever get done quickly.
Answers two and three are both slowly sacrificing detail in favor of brevity. Both are significantly faster than answer one, but have enough detail to be meaningful and answer the question. Depending on the situation, either could be an acceptable answer to Kim’s question.
Answer four is the 0% JPEG of the conversational world. Although it provides some information, it’s not really enough for most purposes, and violates Grice’s Maxims of Quantity and Manner, this time at the other extreme. This answer would likely only frustrate Kim, and would make Pat sound like a Smart-Aleck.
So, in language too, we have to make this compromise. Is detail more important than brevity? What detail should we include? All of these interactions were lossy, and although the degree of loss wasn’t problematic until example four, it’s still vital to keep this in mind when examining language.
So where’s the TIFF?
The biggest difficulty with this metaphor is trying to find out what the TIFF file of Pat’s response is, and does such a thing exist? Although answer one was long and drawn out, there is still some missing detail there that Pat might have known. He didn’t include any clarification of how long it takes to get to the Salon, nor information on other stops that Mom might’ve been planning. No information was given about Mom herself, her manner of getting to the salon, or what else was said before she left. Although you could argue that some of that information might’ve been shared knowledge or knowledge easily assumed by Kim, the fact remains that there is always more that one could say about a subject.
Is our thought the TIFF file of conversation, the singular idea which contains all the detail which we discard when formulating speech? If that’s the case, would it be possible to find or create a “lossless” language? Perhaps this idea of lossless language is what I’m referring to with “High Precision language”.
Maybe it’s even one step further. Maybe the entire sum of our experiences and knowledge formulate one massive TIFF file, and all we do in conversation is crop and JPEG it as is fitting for the context. I kinda hope not, though. A file that big would take forever to open in Photoshop.
Have a question, comment, or concern about this post? Contact me!