What are linguistics foundations of how a message can be on topic?

Question

What I am looking for

Is there a linguistics perspective on a message being "on topic" or "off topic" in some context and/or something inbetween?

Is there even a linguistics definition and research of and on "topicality" and its measurability? Is topicality even a common linguistics term?

Background

What I've observed so far, that for example StackOverflow's on-topic definition is due to the target group of the site being quite large scope "professional and enthusiast programmers, people who write code because they love it."

Then, there are satellite specialized communities like ServerFault, Super User, DevOps, Linux&Unix and Software Development.

Which target groups and domains do these communities address?

• ServerFault: "for managing information technology systems in a business environment".
• Superuser: "for computer enthusiasts and power users".
• DevOps: "for software engineers working on automated testing, continuous delivery, service integration and monitoring, and building SDLC infrastructure."
• LinuxUnix: "for users of Linux, FreeBSD and other Unx-like operating systems.*"
• Software Development: for professionals, academics, and students working within the systems development life cycle.

My findings so far

What I've found so far to learn more about "topicality" and its context for SO/SE Q&A sites is:

• "topicalization" which seems to be something else.
• Google Scholar results are somewhat confusing me in terms that I'm not a linguist so I do not even know which terms are proper to search for in this context.
• For example The good, the bad and their kins: Identifying questions with negative scores in StackOverflow (Aurora et al. 2015) explores on how a machine learning approach can predict "good" questions based on similarity to existing questions. But then, what about questions with new concepts can be asked which are less similar to existing but still on topic?

There is a high-voted SE Meta question with a well-elaborated review of different topics of the above and further related communities: Which computer science / programming Stack Exchange sites do I post on?

Remarkably, a moderator's comment by @CodyGray to this question by has got also quite many upvotes: "I'm not new to SE, and I still don't understand the overlap of topics across multiple sites, so don't feel too bad!" Still, it remains up to the user, experienced SO/SE member or not, to decide which site is best to post on.

Further thoughts

So maybe there is an objective way to define/assess to put a question into context in a measurable way with an assumption that measuring its "topicality" could be a possible approach. Before starting designing some technical solution in terms of computational linguistics and/or data science, I would like to learn more about linguistics view on the problem, even to learn proper terms and find out about related recent research.

Answer 1

No from my knowledge/experience there is nothing like this in Linguistics. At initial glance it is a very complicated phenomenon which requires interpreting the question according to the "topicality rules" of the system, which would have to be defined programmatically somehow because currently there is little to be done in terms of automatically having the computer figure out the meaning of a sentence/question :p. So you would have to first understand (programmatically) the content of the system (the rules for topicality), then understand the question, and if it fits into that rule system. That is I think an unsolved problem with probably little research. They're still working on understanding and summarizing documents, let alone figuring out if something is on topic.