Calle Börstell
LightNews LightNews
banner
cborstell.bsky.social
Calle Börstell
@cborstell.bsky.social
1.7K followers 480 following 220 posts
Linguist at the University of Bergen 🇳🇴

#SignLanguages, #linguistics, #RStats & #dataviz
Posts Replies Media Videos
cborstell.bsky.social
People who use ELAN for research: What is your preferred way of analyzing annotated data?

1) Searching within ELAN
2) Exporting from ELAN and opening as spreadsheet
3) Exporting from ELAN and importing into R (or similar)
4) Other?

#Linguistics
November 25, 2025 at 9:11 PM
Reposted by Calle Börstell
andrew.heiss.phd
I don't know how I've missed it because it's *right on the documentation home page*, but if you use {glue} for nice string interpolation in #rstats and you have {stringr} loaded (likely through the tidyverse), you can use str_glue() instead of glue::glue() or loading library(glue) glue.tidyverse.org
Geordi LaForge rejecting this code library(tidyverse) penguins |> mutate( nice = glue::glue( "{species} on {island} Island ({body_mass} g)" ) ) |> select(nice) #> nice #> 1 Adelie on Torgersen Island (3750 g) #> 2 Adelie on Torgersen Island (3800 g) #> 3 Adelie on Torgersen Island (3250 g) Geordi LaForge approving this code library(tidyverse) penguins |> mutate( nice = str_glue( "{species} on {island} Island ({body_mass} g)" ) ) |> select(nice) #> nice #> 1 Adelie on Torgersen Island (3750 g) #> 2 Adelie on Torgersen Island (3800 g) #> 3 Adelie on Torgersen Island (3250 g)
November 25, 2025 at 6:00 PM
cborstell.bsky.social
This was such a cool map animation of Germany that I had to try to make one for Sweden! But with only ≈2000 towns, so some sparsely populated areas don't show so well. #ggplot2
November 24, 2025 at 2:17 PM
Reposted by Calle Börstell
nataliajagielska.bsky.social
AI slop gracing the cover of Royal Society B. Not only in AI yellow but scientifically nonsensical. Come on. I'm certain human photographs and artworks were ignored to platform ... this.
Royal Society B cover with AI generated image of endocasts.
November 18, 2025 at 10:49 AM
Reposted by Calle Börstell
davidmsidhu.bsky.social
When you think you’ve finally gotten through the last round of reviews.

Reviewer 2:
a picture of yoda with the words no there is another on the bottom
ALT: a picture of yoda with the words no there is another on the bottom
media.tenor.com
November 18, 2025 at 3:24 PM
Reposted by Calle Börstell
mdemeulder.bsky.social
🚀 Call for Papers: #DeafTech2026. We want anything on tech + deaf experiences. Deaf interpreters using captions, deaf animator & game devs, speech-to-text in daily life, sign language AI, VR/AR, media & streaming and more critical/theoretical reflections. Deadline: 8 Dec AOE www.deaftech26.eu/cfp
Call for Papers - Deaf Tech Conference 2026
Call for Papers - Deaf Tech 2026. Submit your research on deaf technology innovations.
www.deaftech26.eu
November 18, 2025 at 1:12 PM
Reposted by Calle Börstell
timoroettger.bsky.social
To all educators out there: We are trying to formulate a criterion for our exams that tackles the vagueness of LLM output and punishes it more strictly. If you recognize the issue, has anyone of you come up with a crisp operationalization of such a criterion. Also grateful for other pointers 🙏
November 14, 2025 at 8:29 AM
cborstell.bsky.social
Smeared charcoal with #ggplot2
A grayscale picture of a bearded man's face in two rows and three columns. Each panel from top left to bottom right gets progressively more smeared looking
November 17, 2025 at 9:59 PM
Reposted by Calle Börstell
fbisnath.bsky.social
#norge #norskforsk
fbisnath.bsky.social Felicia Bisnath @fbisnath.bsky.social · 13d
Hei! Jeg ser etter norsktalere som vil gjøre et eksperiment 🔎. Du vil se bilder og norske ord og si om de er like eller ikke. Eksperimentet tar 10 minutter og du må gjøre det på PC/nettbrett. Velg ett eksperiment og del med andre nordmenn!

🔎 1: tinyurl.com/forsk1
🔎 2: tinyurl.com/forsk2

#norsk
November 17, 2025 at 10:42 AM
cborstell.bsky.social
Elementary, my dear Watson
#TidyTuesday

While the full phrase is never there (see also "Luke, I am your father" & "Play it again, Sam"), you can find "My dear Watson" multiple times across the Sherlock Holmes novels – but which lines? 🔎📖

Code: github.com/borstell/tid...

#r4ds #DataViz #ggplot2
A chart with the title "My dear Watson ..." and subtitle caption "In the four canonical Sherlock Holmes novels, the phrase “My dear Watson” occurs 15 times — 10 times in "The Hound of the Baskervilles"; not even once in "A Study in Scarlet". In the panel below, each square tile (■) represents a line of text — in chronological order — with the highlighted tiles showing the location of the famous phrase in each of the novels." Each highlighted tile representing an occurrence of the phrase is marked with a magnifying glass and a path line tracing the chronological order of phrases across books. Data: {sherlock} via TidyTuesday; Packages: {tidyverse, ggview, marquee}; Visualization: C. Börstell
November 17, 2025 at 9:45 AM
Reposted by Calle Börstell
fbisnath.bsky.social
Hei! Jeg ser etter norsktalere som vil gjøre et eksperiment 🔎. Du vil se bilder og norske ord og si om de er like eller ikke. Eksperimentet tar 10 minutter og du må gjøre det på PC/nettbrett. Velg ett eksperiment og del med andre nordmenn!

🔎 1: tinyurl.com/forsk1
🔎 2: tinyurl.com/forsk2

#norsk
November 12, 2025 at 2:39 PM
cborstell.bsky.social
#TidyTuesday in the style of Chartle showing estimated incidence of TB

Code: github.com/borstell/tid...

#R4DS #DataViz
A line chart in the style of Chartle (a statistics guessing game) on a dark gray background with light gray/white text and lines, showing "Estimated incidence of tuberculosis per 100 000 population". One line is highlighted in coral red, showing the change for Marshall Islands from 2012 to 2023, with a huge surge from around 400/100k to well over 1000/100k in 2018. Data: World Health Organization via {getTBinR} and Sam Abbott via TidyTuesday; Packages: {tidyverse, scales}; Visualization: C. Börstell
November 10, 2025 at 9:48 PM
cborstell.bsky.social
Purrr is great and while I can still struggle with remembering the exact behavior of map functions, {purrr} is definitely becoming a staple for my own coding
Bar chart showing proportion of my R files that call {purrr} functions, showing an increase from 0% in 2020 to 30% in 2025
November 4, 2025 at 2:45 PM
cborstell.bsky.social
So it finally happened. I had to translate an expression from Norwegian because I couldn't retrieve it in Swedish when writing a thing (in Swedish)
November 4, 2025 at 11:25 AM
cborstell.bsky.social
Flint water samples #TidyTuesday

Bootstrapped water samples for estimating proportion of samples with dangerous levels of lead

Code: github.com/borstell/tid...

#R4DS #DataViz #ggplot2
A plot with the title "Lead concentration in Flint water samples in 2015", consisting of a distribution plot (left) and a map of Michigan with Flint marked, with a water pipe coming out of Flint ending in a faucet emoji. The subtitle contains a longer text explaining "In 2015, the Michigan Department of Environment (MDEQ) collected 71 water samples in Flint to evaluate the concentration of poisonous lead in the drinking water. The Lead and Copper Rule posits that if more than 10% of samples are above 15 parts per billion (ppb), action is required by officials. The MDEQ originally excluded two samples with high readings, thus resulting in below-threshold readings overall. Suspicious of the results, a citizen science project coordinated by Prof. Marc Edwards and colleagues at Virginia Tech collected 271 new samples, which pointed to generally higher levels of lead concentration than the official MDEQ data had suggested. The plot below shows the distribution of samples in relation to the 10% threshold when each dataset is resampled 10,000 times using bootstrapping (the MDEQ data bootstrapped both with and without the samples originally excluded). When including the originally excluded samples, both datasets would indicate that the Flint drinking water is likely to contain dangerous levels of lead at a rate which would require official action for public safety. In 2014, changes made to the water supply source in Flint, Michigan resulted in a public health crisis with many residents being exposed to dangerous levels of lead in their drinking water. Data: MDEQ & Virginia Tech in Loux & Gibson (2018) via TidyTuesday | Packages: {tidyverse, ggarrow, ggdist, marquee, patchwork, raturalearth, rsample, scales}| Visualization: C. Börstell"
November 3, 2025 at 3:23 PM
cborstell.bsky.social
library(ggplot2)
ggplot() +
geom_density(data=data.frame(x=rnorm(1e5)),aes(x))+
geom_line(data=data.frame(x =-5:5,y=cos(-5:5)/100),aes(x,y))+
annotate("point",x=c(-.3,0,.3),y=c(.3,.25,.3),size=3) +
ggtitle("Happy Halloween")+
theme_void(base_family="Rubik Wet Paint")

#RStats #ggplot2
A density plot with added eyes and mouth to resemble a ghost and a title saying Happy Halloween in a spooky font
October 31, 2025 at 11:16 AM
cborstell.bsky.social
Maybe Don't Publish In
predatory journals
October 25, 2025 at 5:23 PM
cborstell.bsky.social
Friday night early #TidyTuesday of British Literary Prizes.

Wanted it as a bookshelf, made it into a bookshelf 📚

Code: github.com/borstell/tid...

#R4DS #DataViz
A plot with the title "British Literary Prizes (1991-2022): Proportion of women & non-binary authors among literary prize recipients over time". The plot is designed to resemble a bookcase, with each 5-year time period from 1991 to 2022 featured as a single shelf, and books representing individual prize recipients. Women and non-binary winners are color coded pink-ish and light blue-ish, respectively. Women and non-binary authors were in the minority until 2010–2014 and have constituted about half of the winners since then. Data: Post45 Data Collective via TidyTuesday; Packages: {tidyverse, marquee}; Visualization: C. Börstell.
October 24, 2025 at 10:28 PM
Reposted by Calle Börstell
damiepak.bsky.social
Can you predict when Lou Bega's greatest hit, Mambo No. 5, was released based on the names of all the women mentioned in the song?

Weirdly yes.

yawpr.substack.com/p/project-a-...

#databs #rstats
October 21, 2025 at 8:24 PM
cborstell.bsky.social
Since I have most of my peer reviews saved, I wanted to see some quick stats on the word length of reviews. I suspected they had gotten shorter over time, but it seems less than I thought.
A scatter plot with the title "Number of words in my peer reviews of original and revised article submissions", with points shown either as red squares (original) or gold circles (revised), spanning from late 2017 to late 2025, with about 4 to 8 data points per year and most points falling between 500 and 3000 words (revised manuscript reviews somewhat lower).
October 21, 2025 at 2:16 PM
cborstell.bsky.social
Playing around with subsampling and color tweaking in #RStats and #ggplot2

Ended up with this painting-like rendering of a photo of mine from 2019 🏔🖼
A photo of twin mountains in the distance under a large cloud on an otherwise blue sky, beyond green lands. The image is pixelated in a way that resembles an oil painting.
October 21, 2025 at 12:28 PM
cborstell.bsky.social
UK weather data #TidyTuesday

Temperature changes over time and some individual temperature extremes highlighted on map
🌡🇬🇧

Code: github.com/borstell/tid...

#DataViz ##ggplot2
A line chart showing temperature changes (daily min and max readings) from 22 weather stations across the UK. The line charts include model fit lines showing average temperatures across stations over time (1970 to 2024), showing the estimated average increase at the right margin (about +1.5 degrees Celsius). On the right of the line chart, there is map of the UK showing the location of each weather station. Four of them are marked with annotations showing the highest (Cambridge, 2008) and lowest (Newton Rigg, 2010) monthly means and the hottest January (Yeovilton, 1990) and coldest July (Lerwick, 1993). Data: UK Met Office via TidyTuesday; Packages: {tidyverse, geomtextpath, ggarrow, ggtext, glue, rnaturalearth, scales}; Visualization: C. Börstell.
October 20, 2025 at 7:30 AM
cborstell.bsky.social
New paper with @ryanlepic.bsky.social: doi.org/10.1515/ling....

Our previous research showed that ✨️inherently plural meanings✨️ favor two-handed signs across sign languages, now we see that hearing non-signers also produce more two-handed forms when asked to gesture plural meanings! 👐🏼
#linguistics
Non-signers favor two-handed gestures when expressing inherently plural meanings
Sign languages have been shown to favor two-handed forms to express plural meanings, and in this paper we investigate whether this pattern is similarly found in silent gestures improvised by hearing n...
doi.org
October 16, 2025 at 2:01 PM
cborstell.bsky.social
The real magic is ✨️regex✨️
October 15, 2025 at 6:46 PM