Pinned post

Question for people who make use of the text descriptions for images: If I post a screenshot that includes a lot of text (like I just did), do you prefer a general description of the image, or a full copy of the text? Or is it more contextual (you don't need to know, for example, the full details of my conference submission success page, but it's probably a good idea on a tumblr text meme, so long as it fits)?

I was re-watching the Goat Simulator 3 trailer again and I could've sworn the song was in Simlish.

It's French...

R, data analysis, not linguistics-specific 

For example, here are two F0 plots from a dataset I'm working on at the moment. In order to make all the male sets comparable, I want the lower and upper limits of the y axis to be the same for every graph.

On the left, I have used coord_cartesian(ylim=c(70,160)). This looks like a fairly normal mid-level tone contour. The fact that the creaky one is lower is no surprise, but the difference is important for analysis.

On the right, it's just ylim(70,160). The difference between these two plots is *huge*. And any analysis would be similarly different.

I know a lot of people probably know this. But, I think it's an important PSA for people.

These are both GAM smoothed averages of multiple speakers producing multiple tokens. I don't know if the difference is as stark with other sorts of plots, but good to keep in mind regardless.

Show thread

R, data analysis, not linguistics-specific 

So, I discovered something recently with ggplot2 that some of my fellow phoneticians (and potentially other linguists/data scientists) may be interested to also learn.

If you're doing plots (especially time-plots, e.g. F0 across time) and you want the scales of all your plots to match, using lims(), xlim(), or ylim() merely subsets your data with those limits in mind. It doesn't just set the x/y scales to have those limits.

For many applications this is probably a pretty trivial difference. But if, for example, you're doing data for a bunch of speakers, it can actually change the contours (in some cases, quite severely).

What you want to use, instead is: coord_cartesian(ylim=c(A,B)) or coord_cartesian(xlim=c(A,B))
Where A and B are the limits you want on your scale of choice.

In 48 hours I'm going to be somewhere over the Pacific

The Document of Recovery instructions from the CDC are as clear as mud and a right pain to find on their site

I don't know enough about Linux to really get this, but I figure there'll be some nerds on here that will likely get a chuckle

Commentary from so many horror luminaries throughout the years.
And then they also have ... Corey Taylor? And Chris Jericho?
They very clearly know a lot about horror and are very knowledgeable about the genre. But also kinda stand as the odd ones out.
I don't necessarily think it's bad they're in there, it just seems real weird

Show thread

I've been really enjoying In Search of Darkness 1 & 2. 9 hours of a history of 80s horror? Count me fuck in

Covid, gross 

Some good chunks of lung butter today

Been watching Eli Roth's History of Horror on Shudder, and it's great but Shudder Australia doesn't have the rights to season 2, just seasons 1 & 3.

It's a Shudder show, too. So I have no idea how that's happened.

Just a reminder: If you post about blockchain, you're almost certainly a fuckwit and very likely to cop a block.
Don't be That Guy.

Uber driver is Chinese so I got to practice some. Absolutely delighted I can still have a conversation

Imagine waking up one day and realising you're the spitting image of Kyle Sandilands

There is a guy at the Tote who looks like Kyle fucking Sandilands

Show older

Welcome to thundertoot! A Mastodon Instance for 'straya