Looks like tokenisation will in the future be required before doing anything, according to the warning message:

'kwic.character()' is deprecated. Use 'tokens()' first.

Quarto seems like the future and avoids many of the limitations with base rmarkdown site building, but we don't need to migrate right away - it might be better to wait a bit for more of the process to be streamlined.

Most content will migrate seamlessly, the only major piece of work associated with the move is migrating the styling.

One issue that we need to investigate further is slow pandoc render times for some specific files. This isn't associated with the knitting step, but the pandoc step of turning markdown into html.

Hi,
I am going through the tutorial https://slcladal.github.io/net.html
But there is an error when running

net = network::network(romeo,

•                    directed = FALSE,
  

•                    ignore.eval = FALSE,
  

•                    names.eval = "weights")
  

The error message is:
Error: loops is FALSE, but x contains loops.
The following values are affected:
- x[1, 1:2]
- x[2, 1:2]
- x[7, 1:2]
- x[12, 1:2]

I am running on R4.1.1, would you please fix this error?

So we can apply styling/header changes without doing the analytical work.

Challenge:

• Do this while keeping knit workflows in place

Hi @MartinSchweinberger, in looking through some of the text analysis tutorials, you seem to use "word embedding" to denote the frequencies of a word's contexts. This is a bit jarring to me, where I understand an embedding as a representation in finite, real-valued vector space of a (usually) discrete entity, such as a word. The intention is that distance in the vector space should have meaningful correspondence in the entity space. For word embeddings, this may directly encode the frequencies of its contexts, or may encode a compressed version of that sparse embedding, expressing Harris's distributional hypothesis. Increasingly, however, word embeddings are not simple decompositions of the sparse frequency space, but are discriminatively trained to predict the word's contexts, and may be conditioned on other contextual information to handle polysemy.

Updating both Rmds and htmls to make sure time is shart not European

ariya/phantomjs#15344

phantomjs is broken out of the box on at least recent ubuntu systems - with no active development this is likely to get worse. It would be better to migrate content to a different selenium supported browser.

Can we mix that up the workflow a bit to use less memory?
This is important for two reasons:

1. We shouldn't assume that readers have access to a machine with that memory (apart from colab most online services are going to have much tighter memory limits, and 8GiB of ram is a common spec for a low end laptop).
2. This exceeds the limit of github actions on github hosted runners (7GiB), limiting our options or adding extra work if we start to automate more of the build in future.

1. Links in r examples should point at the new absolute address
2. Any other markdown link should be a relative link