Jerry Bartlett of Content Pro Tech Ltd poses a classification question that brings semantics home in this guest blog post. Enjoy!
Is a hotdog a sandwich? And what the answer got to do with AI or technical communication?
My answer is “No”. Or perhaps, “No, of course a hotdog isn’t a sandwich!” Nor is a burger a sandwich either (but then again, I’m a Brit). But does it cosmically speaking matter?
Well, it turns out it matters because of semantics, and this has everything to do with AI understanding (so to speak) and producing sensible results from your content. Allow me to explain. (Caption and ALT intended.)
Ce n’est pas un sandwich (or is it?)
What is a sandwich anyway?
This hotdog/sandwich question came up in a recent Lingthusiasm podcast – I don’t suppose I need to disentangle that portmanteau for you, except to add that this entertaining pod is produced by two fast-talking and fast-thinking women – Gretchen McCulloch from Canada and Loren Gawne from Australia.
The subtitle of Episode 106 is: The problem with definitions. After some discussion about whether increasingly doubtful things that are labelled sandwiches are really sandwiches, co-host Lauren says, “This debate is actually a really important thing to deal with when you study semantics, which is the meaning of words – definitions.”
Lauren and Gretchen then discuss prototype theory, which if I understand, in linguistic circles, often reduces to an argument about, “Well, it depends what you mean”. Or indeed, how you define things. And that is where we start to take interest as tech communicators, especially in this so-called age of AI <grimace emoji>.
Telling AI in no uncertain terms
One of the jobs of a technical communicator is to define a glossary – a list of the terms relevant to our documentation and their definition. This is mostly for the benefit of our readers, but it is also a way of ensuring the consistency of our company’s brand, IP, area of working, and so on.
The glossary writer works with marketing and product management to define the prototype, to create a parallel with the linguistic argument – the definitive definition, if you like. We decide if we want to call a spade a pneumatic shovel and end all argument.
A glossary definition is important for all the humans involved but is also important for our AI audience. We decide if a hotdog is a sandwich and define it so in our glossaries, terminology databases, and taxonomies – called controlled vocabularies.
As well as defining our terms, we also create relationships between our concepts, and other aspects of our content that show connections that create semantics, so there is no doubt for the AI tool that is “reading”. This goes beyond “mere” definition of meaning. We get to say that a hotdog is a prepared foodstuff that belongs to the category sandwich (or not, as you wish). A calzone is a prepared foodstuff that belongs to the category pizza. A Bombay duck is a dried fish that belongs to the category… well, you define what the category is or what relationship it has to the rest of the things in your universe of content.
The AI tool used to parse your text is “informed” through this set of semantic relation-enhanced statements and is therefore much less likely to recommend you pack an ice cream sandwich in your lunchbox. Or tell you a hotdog is some kind of canine.
Useful links for further research
Items mentioned in the post:
- Listen to the podcast: Lingthusiasm Podcast, Episode 106: Is a hotdog a sandwich the problem with definition
- Transcript of the podcast: Transcript of Lingthusiasm, ep. 106
Want to get in touch with Jerry? You can find him in lots of places:
- LinkedIn profile: https://www.linkedin.com/in/jerry-bartlett-content-pro-tech/
- Content Pro Tech website: https://contentprotech.com/
- Bluesky professional (FWIW): https://bsky.app/profile/jerrybartlett.bsky.social
- Also, he’s a pretty famous beer and whisky blogger and competition judge
