0

I feel like the site is moving to include more programming as a central topic. I feel this shift makes sense.

1. Programming is becoming increasingly ubiquitous, as a common skill akin to reading and writing.

2. Programming is extremely connected to linguistics. The NLP explosion of late has meant that a lot of formerly philological questions for people who read broadly are now computerizable. There are fewer and fewer linguistics questions that do not have an immediate, data-centric programming approach which can generate answers via programs in real time, as opposed to scholastic research methods.

3. Arguably, linguistics was already deeply connected to programming anyway, since natural languages have some level of regularity, and can be analyzed in part algorithmically.

4. It is getting harder to avoid the connection between linguistics and programming given how prevalent the use of linguistic AIs is becoming.

5. Natural language processing should be accommodated by linguists. The computer scientist Edward Dijkstra said that calling computer science “computer science” was akin to calling surgery “knife science” - an incisive point which cuts right to the heart of the matter. Modern computers are arbitrary machines for carrying out procedural processes which can allow people to see how information can be changed to show relationships it has within it. Using programs is a means to an end: analyzing language. Programs are tools, not the focii, but a craftsman must study the tools of their art all the same.

This is my compact justification for that Stack Exchange Linguistics should enable runnable code snippets (“Stack Snippets”).

“If you feel that a particular site should have Stack Snippets, post a feature-request on that site’s meta – if there’s support from the community there, we’ll enable them.”

It would simply mean that any question where small excerpts of code are relevant or helpful in the question body or an answer become runnable with a button where you can see the result. I believe this would benefit programmers and non-programmers alike, since users could press the run button and make use of the result, like a teeny functionality, informational result, interactive diagram, or single-use micro-applet.

If you agree, you may “show support”, in order for Stack Exchange to consider enabling it (it already exists on other sites, see here.)

It would allow me and others to answer a question like “Linguistic Analysis of ChatGPT’s default style of writing” with code they can walk through step by step interactively, like a Jupyter Notebook, to get assistance with both the method and the results of their question. This would be better than providing code to novices who don’t know how to run it; linking them to tutorials off-site on how to do so; or tiresomely trying to explain how to do so via comments. It would also be better than linking to a notebook-type platform off-site, when runnable code could be within the answer.

For example:

Does anyone know what ChatGPT’s characteristic stylistic features are in its writing? How best to analyze that?

Example:

First you start by getting a large language model - try this one - https://mlc.ai/web-llm/.

npm install web-llm

[RUN]

Then you..

*(The code above doesn’t make perfect sense, but it’s an example.)

1
  • 4
    I'm not sure I want to encourage a lot of programming heavy computational linguistics questions here. The programming side of CL probably belongs on other sites. I definitely don't want to see an increase in LLMs here, unless they're sourced from ethical corpora.
    – curiousdannii Mod
    May 24, 2023 at 3:32

1 Answer 1

4

My question is, how many answers would make use of this? When code snippets are posted in answers here, they tend not to be full runnable programs—even in the example you gave. (What would installing a Node package in a web browser even do?)

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .