1

for example, If a question is like

"In place X, how can I go to point A to point B in safest way? should I choose P or Q ?"

The automation should add tag(s) (if available,even if synonym)

"X", "A", "B", "safety" 

to the question.

IMO, If this can be implemented "properly", organizing the site-contents would be far more easier.

what do u think ?

2
  • 3
    It's not simple at all to pick out the words which are names of places. This is a well-known problem within the field of natural language process called "Named entity recognition". – hippietrail Feb 11 '13 at 3:26
  • 1
    This idea is as old as tagging on Stack Exchange and I believe you can find much written about it on meta StackOverflow and the blogs and podcasts. – hippietrail Feb 11 '13 at 3:27
4

This exists as an experimental feature on Meta Stack Overflow and Stack Overflow, perhaps others.

A developer states that it takes about 20,000 questions on the site to get enough material to analyze. This site is still far short of this figure.

On MSO, in my experience, the tag suggester can find one of the main tags but not all of them. On SO the results are worse, it often can't find the most important tag at all.

8

No automated system would ever come close to what humans can do to organize and categorize information. Our tag policy is vague to non-existent in many cases, and if we can't reach a consensus between ourselves, how would an automated system be of any use?

As on any SE site, the tag system is in constant flux, and changes and improvements are made all the time. Besides, the volume of questions on the site is not that big that we can't keep up -- even on a busy day, a single user can (re)tag all new questions in a matter of minutes, if necessary.

4

When I am new to some Stack Exchange site (e.g., this one), I always browse through the available tags to select a few that apply on my question. But, I always doubt if I really have the appropriate once, and then rely on the community to correct them when they're wrong, or when there are some to be added.

A feature like 'suggested tags', would be awesome, but I wonder if it could do more than just suggest the obvious ones, which the author probably already thought of. In your example, I think "X", and probably "A" and "B" can easily be recognized, but "safety" depends on the context, and it thus more difficult to select.

If you are a experienced user, and read the question anyway, will re-tag appropriately right away, to make sure the contents stay organized.

As full automation is not possible, I would say to just let the community organize it.

4
  • see the first comment on the question. It'd be great, but it's a huge area of computer science to work out natural language parsing, and determine what the person is talking about. They mention great britain (missing caps) - did they just say 'great' as an adjective, or was it a reference to the island of 'Great Britain'?. Similar for new zealand, south africa (is it the country or south in the continent) and more. – Mark Mayo Dec 27 '13 at 12:34
  • @MarkMayo Well, if you make it sophisticated, yes. But here you only have to check against a database of simple tags (200 maybe?). That is not difficult. Also, for the typo's there already is a solution: en.wikipedia.org/wiki/Levenshtein_distance – Bernhard Dec 27 '13 at 13:01
  • Indeed, I've used the Levenshtein distance in past projects, but it still doesn't help with context issues. It's a problem I'd love to help solve for SE, but I feel there's a Masters or Phd level of effort work in the problem ;) – Mark Mayo Dec 27 '13 at 13:05
  • Of course, if you don't mention the country at all, it will be difficult :) – Bernhard Dec 27 '13 at 13:09

You must log in to answer this question.

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