up vote 31 down vote favorite
1

Proposal: Computer Science (Non-Programming), Natural Language Processing and Computational Linguistics (NLP/CL), Machine Learning, Artificial Intelligence & Robotics

Should we merge the various Computer Science proposals. It seems to me that a lot of these sites on a specific topic gets very little traffic, and the AI proposal was closed in part for lack of activity.

It seems to me that it would be better to have one site than having many different smaller sites about a specific topic in Computer Science, similar to how StackOverflow does not consist of many smaller sites, one for Java, one for C#, etc.

link
2  
I suggest that you tag this proposal too - Artificial Intelligence & Robotics. – vishvAs vAsuki Nov 22 '11 at 20:09
1  
Crypto seems to have trouble in beta. Is somebody active there who could suggest moving their community to CS? In general, all the other proposals linked here are not yet established. Is there a process to decide "This is not going to work out" (meaning we could suggest a merge) or are proposals just left hanging? – Raphael Dec 1 '11 at 10:25
Computational Science looks fine, I guess, considering they have only been in beta for five days. – Raphael Dec 5 '11 at 7:58
Also "Computer Vision" proposal that I just spotted is probably included here – Merlin Dec 20 '11 at 21:45
I just realised that there is a Computer Science Education proposal. As is has been in definition stage for quite some time and can clearly be seen as subset of the CS proposal, a "merge" might be in order. – Raphael Jan 1 at 13:25
I did top level searches on some ML/AI terms of interest to me and found that biologically-inspired algorithms (GA, GP, EC, PSO, AVO) get talked about mostly in cstheory.stackexchange.com, and the more stats-based techniques (NN, SVM, as well as the term 'machine learning') are mostly in stats.stackexchange.com. Other sites that commonly come up include programmers, quant & scicomp. – Graham Jones Jan 12 at 11:52
My personal preference would be for the broad range of AI/ML algorithms to be covered in one site, both bio-inspired and maths/stats-based, as well as ML/AI applications: computer vision, robotics, system identification, game playing, control, etc. – Graham Jones Jan 12 at 11:56
[status-deferred] until one of the CS communities stands up united to do this. See my answer below. – Robert Cartaino♦ Feb 6 at 21:58
Also "Computer Architecture and Organization" [area51.stackexchange.com/proposals/19424/…. – espertus Feb 21 at 9:09
feedback

10 Answers

up vote 19 down vote

As Joel Spolsky wrote: “the right size [of an SE site] might be somewhere around the size of a university department”. So yes, a computer science site would be the right size, rather than specialized subfields.

The limitation is for cross-subject specialties. For example the Cryptography site has a very narrow scope, but it wouldn't exactly work as a part of Security (a lot of questions are too mathy) nor of Math (a lot of questions are too close to implementation). Mind you, it might have worked as a subset of Computer Science. If there had been one, that is. So, machine learning: I think that can work on CS. NLP, maybe not so much, because most computer science people aren't used to working with linguists and vice versa; so I think it would work better as its own site.

link
I agree that "Machine learning" might work. Computer Science seems way too broad, although I really think NLP and ML could be the same SE site! – sandos Jan 4 at 14:11
@sandos The Math and Physics sites are working well. Computer science is the right size. Area 51 tends to produce sites that are too small to sustain a community; we shouldn't try to follow that trend. – Gilles Jan 4 at 21:02
isn't Stackoverflow a subset of Computer Science? – bizso09 Jan 16 at 17:15
6  
@bizso09 No, not any more than Motor Vehicle Maintenance and Repair is a subset of Physics or than English Language and Usage is a subset of Linguistics. – Gilles Jan 16 at 18:27
feedback
up vote 9 down vote

I don't think a distinguished merge is necessarily needed. If there is a specialised community, people should go there with their questions. If not, they choose the next appropriate one, which in case of AI might be either one of CS, Math or Stats, depending on the question.

What we could do is approach proposals of CS subfields that struggle (if existent) and ask the supporters to join forces.

link
I like this idea. Certainly, if a topic begins to garner a lot of traffic, that would be a good time to create a new community. – Paul Oct 5 '11 at 8:51
Graph theory questions are being answered on cstheory.SE even though there are complexity theorists and others around. I think as long as the level is more or less homogenous, different fields can share one site. Overlaps only increase a community's helpfulness. – Raphael Dec 4 '11 at 16:09
feedback
up vote 6 down vote

There are only a few substantial paradigms in programming; most expert programmers know a couple and can answer Java questions despite being a C# expert, and so on.

The things you can do with those languages, though, is unlimited. Experts in Computer Science don't need to know anything about Machine Learning or NLP; their expertise may be in compilers, or parallel programming, or something else. These areas have plenty of room for Q&A.

I didn't participate in the AI site but I understand that it was actually too broad. Beginners asked uninteresting beginner questions about everything and there was no expertise gathered in any particular area, so the site failed to generate any interest.

link
1  
I totally agree with you. I also think that if a Site is too broad it won't attract the 'real experts'. – eowl Dec 7 '11 at 12:37
2  
@Matthew Read There are only a few substantial paradigms in programming; most expert programmers know a couple and can answer Java questions despite being a C# expert, and so on. That sounds like something a computer scientist would say, not a programmer. Many questions are library specific so that it can't even be answered by just another Java user, but only by another user of that particular library. In that sense Solr or Hibernate questions are as niche as NLP or computer algebra or theorem proving. – ArtB Dec 9 '11 at 15:56
@ArtB I didn't say all specific library questions in C-like languages are answerable by all programmers who use a C-like language, but with that said, it takes years to become an expert in a niche area of CS. Most libraries are picked up orders of magnitude more quickly, and you're much more likely to find an expert in two libraries than you are to find someone with two PhDs for distinct areas of CS (for example). – Matthew Read Dec 9 '11 at 16:03
1  
I'd say it takes about a year to become good with a given library and I think with year's worth of work/study of NLP or machine learning you could answer most peoples' questions. Sure, not the experts' obviously, but most questions. Yes, I am assuming, that most question come from people beginning to use a library/technique/area-of-study or up to novices. – ArtB Dec 9 '11 at 21:46
feedback
up vote 4 down vote

I would second Giles. However, I would suggest that there be no sub-fields at all.

Having specialized fields won't be a good idea. They get too specific. Apart from that nowadays, research and technology is getting overlapped among the fields of CS.

So, a guy doing research in say systems, might need to know something about machine learning or about networks depending on his research. Having different sites for different areas would prevent people from different areas to interact that easy.

Also, there are a lot of tools, concepts and algorithms that are used by not just one, but several departments. Having a unified site, would encourage interaction of people on this stuff, even if they belong to different areas.

link
feedback
up vote 2 down vote

Having checked out the first few pages of proposals here on area51, it does look like there are a lot of proposals that could be merged. There is Artificial Intelligence & Robotics, Machine Learning, Robotics Research, Computer Vision and Natural Language Processing and Computational Linguistics that all could be consolidated under one SE site. That's five different sites, when we could have just one. By consolidating, not only will we save effort getting into beta & beyond, but it would be helpful to have one consolidated resource instead of five related but different ones.

link
I agree in principle, but you won't get the specialists to agree. The promise of a specialised community is all too attractive, and those proposals do not look dead at all. – Raphael Dec 15 '11 at 10:09
feedback
up vote 1 down vote

Why duplicate effort - there is already another AI site .

link
6  
That's not a Stack Exchange site, it's an OSQA site, admitedly similar but lacking the reliability and service level that SE sites can provide. – Merlin Nov 30 '11 at 18:03
feedback
up vote 1 down vote

I whole-heartedly agree that there is good reason to unify these related subjects. I find it pointless that there are so many highly specialist subjects being proposed when such a large audience is required to sustain a site throughout its probationary period.

The question is how do we proceed.

link
2  
"Beta Q&A site for scientists doing science by heavy computation" -- while overlapping literally every science and engineering field nowadays, including computer science, I think computer science and computational science are different enough to justify different sites. The similarity in names can be misleading, I guess. I would expect mainly computer scientists to end up on computer science SE, but people from all fields that use computations on computational science SE. Very different communities. – Raphael Dec 4 '11 at 16:11
feedback
up vote 0 down vote

Creating a combined "Computer Science" site has been attempted on many occasions. The larger community seems to want it, but it is almost always met with contempt and disdain by those participating in the individual subjects.

Area 51 does not handle this very well, so we've been going forward with these individual sites (Cryptography, Computational Science, CSTheory (merge rejected), AI (closed), and now Machine Learning) until one of the Computer Science communities stands up and has a cohesive conversation and comes to a consensus about creating a more comprehensive Computer Science site.

link
1  
Note the same answer over here with some comments. – Raphael Feb 8 at 11:57
feedback
up vote -3 down vote

I think something like academic divisions can be used. Or something like Wikipedia divisioning. Anyway Computer Science is too broad, it includes to many topics. I think the following areas can be marked out:

for general methods:

  • Theoretical computer science - for abstract questions like graph theory

  • Machine Learning, Pattern Recognition, Data Mining - for general methods

    for applied peculiarities:

  • Computational science - for Numerical methods
  • Information science - for Knowledge Representation and Natural Language Processing
  • Robotics
link
4  
Mathematics and Physics work on StackExchange, why should CS not work? Note that your classification leaves out lots of subfields. – Raphael Dec 11 '11 at 16:07
feedback
up vote -4 down vote

proposal: attack on hard open comp sci/math problems

suggest this might fulfill much of elsar's answer on merging Artificial Intelligence & Robotics, Machine Learning, Robotics Research, Computer Vision and Natural Language Processing and Computational Linguistics

I read joels blog post above about separate vs merged sites & show how it fits the bill of questions & directions that are currently being somewhat aggressively rejected on the relatively new cstheory stackexchange.

that site is also suffering a dropoff in question volume and a recent spike in closed-to-open question ratio.

link
3  
CSTheory is 1½ years old, that's pretty mature in web years. From its early days, it's been for research-level theoretical computer science; questions are rejected when they don't fit the site's topic, like everywhere else. – Gilles Jan 20 at 0:56
3  
The linked proposal is not thought through and appears to be the result of the author being disappointed in the way cstheory.SE (and SE in general) are run. – Raphael Jan 20 at 9:19
dont know whats not "thought through" in the proposal. hope to hear specific reactions from anyone on the proposal. it is true Im initially disappointed in cstheory.SE's heavy handed moderation policy which imho may be more extreme than other stack exchange sites. as far as stack exchange I think its very cool software, possibly a "gamerchanger" esp for collaborative science & have said so elsewhere. see linked proposal/questions – vzn Jan 22 at 0:05
feedback

You must log in to answer this question.

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