Linguists have always been interested in reconstructing languages from long ago.

By looking at how languages are now, we can work backwards to see how they must have been in the past. But now computer scientists are creating a kind of linguistic time machine that can do the job of reconstruction automatically. How does it work?

Linguist Daniel Midgley steps into the temporal vortex on this episode of Talk the Talk.

Listen to this episode

Or subscribe via iTunes
Click here for more options: How to listen to Talk the Talk

Language reconstruction is one of the dark linguistic arts, but this time Ben and I are getting into it. It’s like going back in time, deciding what early languages must have been like by looking at what languages are like now. So first, we talk about how language reconstruction works, and then we look at a new project where people are getting computers to do the work.

I hate to say this, but as is so often the case in linguistics, big progress is being made by non-linguists — engineers and computer scientists! Linguists sometimes grump that the engineers aren’t familiar enough with the actual work of language reconstruction, but I love the idea of taking linguistic tasks and making them tractable for a computational treatment.

Even though this is kind of dense subject matter, I think we made it interesting. Thanks to Ben, of course.

Show notes

The press release

Another couple of articles

and the actual paper.

Here’s where they got their data from: The Austronesian Basic Vocabulary Database. Have a look!

Check out some reconstructed data from Proto-Indo-European

‘Tongue’ in English

Saussure’s ideas were verified by the discovery of Hittite.