12

Proposal: Data Science

The data science looks like an overarching umbrella proposal for stackexchange that covers too wide an array of specific fields:

  • Natural Language Processing
  • Big data
  • Information Retrieval
  • Database handling
  • Machine Learning
  • Distributed Computing / High Performance Computing
  • ...

It sounds like a good idea to group all these up into one pipeline but considering the intricate features of each of the fields it cover, it seems unlikely to get enough expertise on answering the questions. Moreover each of the fields that "Data Science" covers includes specific subfields that varies vastly.

I can see that different from CrossValidate (https://stats.stackexchange.com/), this proposal looks more at the technology of data processing rather than the math of data processing. But is there a necessity to alienate technological difficulties in data processing from CrossValidate?

Then again, stackoverflow seems to function very well with different people coding in different languages and also in different domain and the thing that links everyone together is "computing". Using the same ideology, linking all the fields together just because it's about "data" processing is similar in the train of thoughts.

So the ultimate question:

Data Science proposal, an over-ambitious proposal?

or a brilliant attempt like stackoverflow is to "computing", data_science is to "data processing"?

18

I would say that it is no more broad than the topics of many of the most successful StackExchange sites, such as StackOverflow (computer programming) or MathOverflow (mathematics). It seems to me that the biggest problem for a fledgling StackExchange site is attracting a critical mass of users. If the focus is too narrow, not enough people will join, or they'll join but not be active because they're trying to keep up with bigdata.stackexchange.com, machinelearning.stackexchange.com, distributedcomputing.stackexchange.com, etc.

In my opinion, it makes sense to combine sub-topics into a single StackExchange site so long as the majority of the experts it attracts are likely to have some expertise in multiple sub-topics. Or more specifically, the union of the sets of expertises needs to span the set of topics. If you attract a set of experts who know about both "big data" and "machine learning", and another who know about "big data" and "parallel computing" and another who know about "machine learning" and "natural language processing", then these are all related and overlapping topics that draw from a shared pool of experts. Obviously, there might be a few people in that set who also happen to know about medieval European history or some such unrelated thing, but there isn't going to be a strong correlation between knowledge of those two topics.

I certainly don't think it's necessary that anyone have expertise that extends to every subtopic of the site. The design of the stackexchange sites, particularly the use of tags, makes it easy to focus on only the topics that interest you or that you have knowledge of.

Another good criterion (again, which I am completely making up on the spot) is that there is a community who identify with the topic. Gardening might be a good StackExchange topic, because there are people who consider themselves "gardeners" who are likely to want to join if they hear about such a site. Likewise, "mathematicians" or "software developers" are groups that people identify themselves as belonging to. Whereas even having an interest in both physics and Greek mythology, I'm not likely to seek out ThingsWrittenWithLotsOfGreekLetters.StackExchange.com, because it's artificially conflating unrelated interests of mine. Particularly with data scientist emerging as a recognized profession, and with schools starting to offer degrees in data science, it makes sense as a site people would want to participate in.

4

My opinion is similar to Tim's one. Besides managing to find a critical mass (which of course eases the process of creation for the group) I think that the real bonus in a proposal is to put together similar things, even if people are not expert in several of its sub-fields. What really matters is to know that the fields are sufficiently close that (albeit with a little effort) it is possible to understand questions and answers. This would be the main difference with ThingsWrittenWithLotsOfGreekLetters, at least IMNHO: I for once would benefit in learning things in my own sub-field, but explained from a different point of view.

1

This Data Science proposal is on the lines of http://www.scoop.it/t/eedsp?page=1 where you get to know all the happenings/technology/internals/architecture/news in the world of Data Science spanning multiple areas like Information Retrieval, Machine Learning, Distributed Computing, Programming languages (more suited to Data Science Applications), Analytics, Statistics and so on. Please go through it, it is indeed very interesting.

  • +1 for the scoop link =) – alvas Sep 26 '13 at 10:15

You must log in to answer this question.

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