Amazon’s almost-tags Marc Hedlund 15 Apr 2005

11 comments Latest by name

Amazon has added concordances and statistics about the text of a book to their site (hat tip to Nat for the pointer; here's a definition of concordance if you want one, as I did). The concordances use Flickr's size-for-frequency design to show which words are the most popular within a book.

Unfortunately the similarity stops at the presentation -- the concordance looks like a set of tags, but isn't; you can't find other books with the same frequent words. (You can use Amazon's other cool new feature, "Statistically Improbable Phrases," to find books with similar language, as long as the usage is improbable.) It would also be fun to browse across books with common words used frequently in their customer reviews ("hilarious" and "enraging" would both be interesting to see; especially in combination!).

All of this makes me think a search engine for tags isn't as crazy as I'd first thought it was when Technorati launched theirs. What is the meaning of this word within the applications of the web?

11 comments so far (Jump to latest)

Jason Hummel 15 Apr 05

I think the tag specific search engine is the next logical evolutionary step in Web content. If you think about it, all we’re doing is redesigning our ages-old library system, but this time - everything is user contributed. First we had blogs/Web sites (books), then we had RSS/Aggregators to bring all the content together (Library), Then we had Tags to classify the content (Dewey Decimal). The next step, you would think, would be some sort of helper system, like a librarian or a card catelog. I think a tag specific search engine fills this niche quite well.

What interests me is, Ok what’s next? If we really are simply updating old concepts for a new content delivery system, what are we still missing? Figure that out, and you might be the one to launch the next big thing.

Beau Hartshorne 15 Apr 05

Check out this great talk by Clay Shirky on this topic:

http://www.itconversations.com/shows/detail470.html

Jeni 15 Apr 05

I keep hoping that Amazon will eventually implement tags. It seems a natural route for them - instead of using an arbitrary classification, they instead have classification developed by users. It works well for del.icio.us and Flickr (sometimes in unexpected ways), why not Amazon? Likewise, a search engine that returned results ranked in part by tag concordance could be interesting.

I think tags haven’t quite “clicked,” as it were, with the general populace. A great deal of classification systems seem to be mostly singular in nature, so it doesn’t come naturally that an orange could be classified under “round,” “color:orange,” “fruit,” “citrus,” etc. I struggled with that for awhile when I first started using del.icio.us.

mp 15 Apr 05

I work for a reasonably large online book e-commerce site (not Amazon). After explaining and showing the concept of tags to the management, the company has adopted tags as part of the product strategy. The idea is to allow both user assigned tags and company assigned tags to every piece of inventory — books, content pages, reviews, forum postings and so on - much like Technorati et. al. So I do think tags will play a role in e-commerce sites with large and varied inventories.

Jamie 15 Apr 05

Jason, I am really interested in the points you bring up in your comments. I think a librarian/tag based search engine can only take you so far. With all the content being generated (blogs, podcasts, etc) there is so much to sift through. Advertising (or better yet Paid Content Placement) has a major role to play in all of this … but not for big business. Someone or something that can effectively market all this self-published content and get independent bloggers/podcasters more “eyeballs” and “ears” will make out very very well I believe. Marketing is the missing part in the whole Web content equation.

Sorry to diverge from the original topic at hand.

Jeff 15 Apr 05

I also work in e-commerce and I fully expect tags to play a bigger role in the coming months as a way of relating products, it makes perfect sense, I love the way that Amazon can show me books that others have purchased, but I’d to be able to find books that others have group together by topic, by content, author, etc.

The challenge with tags is that they are susceptible to spam, just as blogs have been hit with comment and trackback spam, in some ways I’m surprised this hasn’t become a problem for del.icio.us yet. There’s nothing stopping me from adding links to the site with tags that are popular, but inaccurate….

Su 16 Apr 05

I’m sorry, but isn’t this is a complaint looking for a problem? They’re not (almost-)tags; they’re not meant to be. It’s a “concordance”[1]. Concordances don’t refer to materials outside what is being indexed. It’s contrary to the point.

Saying that the similarity stops at the presentation seems to imply it should go further.
More deeply, it assumes that a strictly visual similarity somehow follows through to a conceptual connection[2]. And even that first requires that you ignore both sites clearly stating their purpose. I dunno. I’d be more accepting of this if they were just mystery boxes with words in them. But I see an important difference between making a suggestion or wondering about a variation on something, and suggesting that it’s failed to do something fundamentally opposed to what it actually is.


[1] That box doesn’t even begin to approach being a concordance, which has a very specific definition. If it’s not comprehensive, it’s just a list of words fitting whatever criteria produced them(in this case the 100 most common).
I’ll at least concede that they seem to be using the word as a repurposed label rather than trying to redefine it, which might account for their phrasing that “Concordance is…” rather than “A concordance is…”

[2] Ain’t necessarily so. Not to invoke Godwin, but let’s talk swastikas. It’s obviously an extreme example, but it’s also got almost no room for ambiguity.

name 06 Feb 06

testing