Ask HN: Alternative to the tag cloud? | Hacker News Reader

Ask HN: Alternative to the tag cloud?

16 pointshelium11 comments

I'm building a website that will list a (hopefully) large amount user created items indicative on their relative importance.

I have considered using the ubiquitous tag cloud for this, but I find it a bit blasé. In my opinion, it's also not very approachable for non-techies.

What are the alternatives?

View on HN

What are the items? Are they best represented by single words or short phrases? (If not, I don't see how you could use a tag cloud anyway.) What does "importance" actually mean? What other characteristics of the items might the users care about? Are there any that you can usefully use to arrange the items in whatever sort of display you choose, putting "similar" or "related" ones near to one another?

Some options:

Big list, starting with most important. Paginate the list if it's too big, or something. Very, very old-school, but worth considering briefly. Will be less unbearable if you can offer ways to filter the list; the same probably applies to all the other options too.

Tag cloud (words with size proportional to importance). Can also manipulate the ordering of tags in the cloud -- alphabetical is standard, but you may have other ideas.

Image cloud (images with size proportional to importance). Can also manipulate the arrangement of the images, draw lines between "related" items, etc.

Random selection of items (selecting few enough not to be intimidating, but maybe give a way to show more), more important items selected with higher probability, plus a button or something to make a new selection.

Find a way to categorize the items. Order categories by total importance in each category. Allow the user to drill down into categories (which may be nested, if appropriate).

Diagram showing a box or something for each item, no text or images or anything so all you see on looking at the diagram is the sizes; show information about the box currently under the mouse pointer. Will work better if you can arrange the items meaningfully, but probably won't work well whatever you do.

Arrange in two dimensions by some characteristics other than importance. Show items in some subset of this 2-dimensional space, choosing the subset by importance and its size so as not to have too many overlapping items. Allow zooming, panning and maybe adjusting of the permitted density. (Think of it as a map with items for cities/towns/etc.)

Have a text box with autocompletion, with suggestions filtered by importance and a fixed target number of suggestions.

Filtering mechanisms:

By importance. By name. By characteristics of the actual items (can't suggest what since you've said nothing about what they are). By when the items were added/changed. (You could have a slider thingy so users can watch how the world has changed over time.) By similarity/relatedness to a particular item or set of items that the user has selected already somehow.

The items represent dynamically defined category hyperlinks. These will all be all be top-level categories that are not nested. They will be enforced to be short single words.

They might be loosely related to each other, although I'm not sure yet.

Importance rating will be dead simple, it will be based on page views in a specific time frame.

Reading the comments has made me think about this some more.

The most frequently used categories should be the real easy to find, more obscure/rarely used might need filtering or searching and should actually be difficult to find, thus discouraging their use and lowering their importance. Hmm... actually it sounds like I'm describing HN.

"The most frequently used categories" in appropriate time-frame, say, last two weeks. Maybe give some random members to the mix so people discover new areas and build up (depends on project of course). Do you remember Netscape's "What's New?" and "What's Cool?" meme?

Here's a discussion on metafilter: http://ask.metafilter.com/56228/Alternatives-to-a-tag-cloud

You can find more experimental UI elements at infographics sites and various visualization lab sites: http://infosthetics.com/ http://labs.digg.com/ http://sandbox.yahoo.com/Experiments (it used to be much better before they cut a bunch of projects) http://www.etsy.com/buy.php

I understand that you're probably looking for an established alternative; however, most solutions need to be appropriate to the situation. Feel free to email me and maybe I can help you brainstorm.

I'd recommend this Edward Tufte book for visual data design, it's very good:

http://www.amazon.com/Visual-Display-Quantitative-Informatio...

Also, FlowingData is a great site for data visualization: http://flowingdata.com/

Instead of differentiating by font size you could differentiate by color intensity (e. g. like in http://www.artlebedev.ru/everything/clouds/linear/ )

Consider using hand-picked instead of generated aggregates.

For example, on a picture sharing site, showing me a mix of other users' latest pics is just noise. But if an editor creates a album of "Old Mexicans Smoking Cigars", that sounds interesting.

Note: anything with a lot of information visibly present in it is "not very approachable for non-techies", so filtering is probably at least as important a problem as presentation of the filtered data.

I prefer a ranked list, one tag per row. Also draw bar graph next to each item to visually show the difference in magnitude between the ranks.

I've always kinda hated tag clouds.

Can you give us a sample of what the data looks like?

Reply on news.ycombinator.com