Hacker News Reader

(Disclaimer: I'm that friend)

I'm pretty conservative about this, since it's hard to completely oversee the influence it has on a community (Slashdot's tag system comes to mind). While the need of categories/tags is already illustrated with the "Ask HN:" in this title, it does raise a few concerns:

- it puts an extra burden on the submitter, do we really want that ?

- is there enough content on HN to justify a category system at all ?

- what if someone submits content in the wrong category, how should we deal with that? should the mods deal with this, thus adding an extra burden on the mods too?

- what if an article belongs to multiple categories?

All in all, doesn't it create more problems than it actually solves?

As the problem is that original submitters can't always be bothered providing a descriptive title, why would they be any more likely to bother providing a suitable tag?

There is an external implementation of tags, although most likely it hasn't taken off because no one uses it, and no one uses it because it hasn't taken off.

http://news.ycombinator.com/item?id=1069726

PG is talking about having multiple flags for items, but that's about appropriateness, not about the semantics of the content:

http://news.ycombinator.com/item?id=1006589

The problem is always getting the elvel of complexity right. Allowing free-form tags makes it impossible to search for what you want, and makes it hard to decide what tags to make up and apply. Having a limited collection makes it impossible for some submissions to decide what tags to apply.

It's hard, but not impossible. It would also prevent me from seeing some of the things I've really appreciated. I would tend to skim over things that don't appear relevant, and sometimes those are the very things that have engaged my interest and expanded my horizons.

It's like with these dating agencies, or friend suggesting algorithms. I don't want to meet people with whom I share an interset - I want to meet people who are interesting.

Solve that, execute well, and you'll make a fortune.

Conceptually, what should tags denote? At some level, I'd argue that tags are solely to give the reader a better idea of subject category. Do others agree?

I ask because (and especially on a discussion site like HN), usage of (free-form) tags to editorialize is inevitable without heavy moderator monitoring. Would the tags 'yes' 'godyes' and 'hellyes' on a link titled "Is Apple Evil?" add anything other than gasoline to a flame full discussion? (Not that flamebait tends to last long on the frontpage.)

Are tags really just for the disillusioned to block all stories tagged with 'ipad'?

> At some level, I'd argue that tags are solely to give the reader a better idea of subject category. Do others agree?

I want something to help me find "interesting to me". If tags are "solely ... subject category", they can't be very useful because subject is neither necessary nor sufficient when it comes to interesting.

Here's a test - are you actually interested in everything in any subject? Are you ever interested in something in a new subject or a subject that you typically don't care about? If your answers are "no, yes", subject category information isn't enough for you.

I don't really care so much about the category of the post, as in programming/startups/CSS/Apple. I'm more interested in knowing what kind of post it is, as in pdf/magazine article/blog article/video etc.

One possible way "to see at glance what a link is about" is to make an automatic term extraction of the link. A small box could be included (I know this could overload a bit the current nice and minimalistic interface) showing the inverse frequency of the most popular terms in the retrieved link using a kind of TF-IDF weight algorithm.

I suppose a "stop words" list would be required along with a stemming algorithm to limit the tokens/"words" with less semantic meaning. In this case, the work can be done while the submitter is adding the link and the software proposes the extracted list automatically. The submitter can alter the list proposed to improve the meaning of terms extracted.

It may well be a viable solution, but I am not convinced that the problem of "the ability to see at glance what a link is about" exists for everyone. Not for me, anyway.

In terms of "anything that gratifies one's intellectual curiosity", the variety that I find by clicking through a high-rated item with an obscure or mystifying headline is welcome.

A tagging system may help me to pre-filter, but by applying my own filters instead of trusting the HN community filter, I would no doubt lose the surprises and intellectual gratification. Some of my favourite items have come through not knowing what the link is about, but clicking anyway.

I agree with each paragraph above. Generally, I consider the quality of most headlines to be sufficient, and in cases where it is not, I still have a second option - click on the comments first, which will give me enough suggestions whether it is worth to visit the site.

Moreover, I surely wouldn't want to pre-filter according to the tags, because one of the main reasons I like HN is that I am confronted with topics, which are not my primary interest, but allows me to expand my thinking to new areas.

To me the viable solution is to sometimes look at the comments to establish whether I should read it.

Sometimes I know it's of interest so I go to where the link takes me, but I just found that often I get a much better summary by someone here at HN than from the url in question.

I always scan the comments first. I'm usually much more interested in the discussion than the original post.

It would be possible to get a classification of the link through opencalais pretty easily (don't know about the legalities about how that service can be used though). I suppose it's mostly useful to be able to browse hacker news by category.

Example: the post "'Chang' from Google accidentally publishes TechCrunch post on Blogger Buzz"

Topic:

"Technology Internet"

Social Tags:

Technology Internet, Yahoo!, Computing, World Wide Web, Yahoo! Mail, LinkedIn Open Networker, AOL, Facebook, Carol Bartz, Communication

Entities: Company Facebook Inc [36%] Time Warner Inc. [30%] Twitter Inc [10%] Yahoo! Inc. [79%] Country United States [26%] Industry Term advertising industry conference [32%] analytics tools [19%] display advertising tools [31%] finance chat rooms [12%] online marketing [6%] Person Carol Bartz [80%] Comscore [26%] Position CEO [32%] Published Medium Advertising Age [29%]

Most of the comments so far have focused on the fact that tagging may benefit readers before clicking through on a post.

I actually think the benefits of tagging would biggest after submission. Namely:

- If you could view articles by tag it would help new users coming to HN view what might have been said before about a given topic

- It would make mashups like this a heck of a lot easier: http://www.gabrielweinberg.com/startupswiki/Ask_YC_Archive

- It would make searching for relevant historical discussions easier. There is so much good content on HN that is effectively lost

Some suggestions for an implementation:

- Make the tags freeform, ala Delicious, so as not to box things in to strict categories

- Bake it into the main site itself - http://hntags.com doesn't work because that's not where we hang out.

- Have the ability to view all articles with a given tag

- To relieve the burden on the official mods give users above a certain karma level the ability to apply or alter tags on submissions

Ask HN: Tags at Hacker News