User talk:Dria/On Tagging

Great! It would play nice with my plan to improve Bayesian to handle multistate. --gandalf 12:01, 15 Oct 2005 (PDT)

Folders (or files) hierarchy aim to be deprecated, for sorting folders, images or bookmarks. I think that the futur is the use of Metadata/category with metadatabot & "browsers by metada" and especially with the help of XPTC core 1.0 technology (aka ~= XMP). http://www.iptc.org/IPTC4XMP/ For me the benefit of this metadata framework, is the use of Dublin Core standard (used also by standards and catalogue coordinator, librarian...), XML RDF standard, UTF-8 standard, open & royalty's free adaptability and portability... (XMP is used in MacOSX search, Adobe's products etc.. for images and files) -- ginkyo 07:56, 16 Oct 2005 (PDT)

Existing bookmarks

How would you import existing bookmarks, and educate the user on switching pardagims entirely. Tagging works because it's to the point, but any auto-tagging done by importing the bookmarks would create a horrible mess and hugely unfriendly to the user who may (or may not) have had their bookamrks previously in an order they were happy with.

That's a really good point. Another person mentioned that web pages often also have really lousy titles, so users would have to be able to modify the title as well as the tags. This article really is just a brainstorm of ideas, and not a full-fledged design, by any means. -- Dria 12:25, 16 Oct 2005 (PDT)

Perhaps the tree-tagging can limit this effect, because the import would result in 'tagification' of the bookmarks tree. The user will discover later that he can give multiple position to a single bookmark. Carmine 06:26, 8 Nov 2005 (PST)

Epiphany

You should try Epiphany, the Gnome official web browser, which is based on gecko. It already have some of the features you propose : It tag bookmark and it's has type ahead search on you bookmark (in the location bar). It may help you to get a better idea of what you exactly want. --Staz 21:46, 16 Oct (EST)

Possibly a good suggestion, but I use OS X. -- Dria 19:36, 16 Oct 2005 (PDT)

Tree-Tagging

One of the falws of tagging is caused by the homonymy of many words. For example, when I tag a link 'Quebec', is it about the town of Quebec or of the province ? Or for the word 'France' ? The first name or the Country ? Having a tag tree will help to solve thoses problems.

Architecture / 2

Tags are popular because they're easy.

Don't try semantic web here, putting rich information in it, like hierarchies or something. Tags have their charms because the user knows if France is a name or a country or both. And can live with it.

To me, having smart algorithms that assist when tagging (don't forget rel="tag", btw) is one half of the trick, the other half is getting back to the URL. I definitly need to browse with Flock a bit to see what they're doing, dria's proposing something really close. Though they still have Favorites, and not just tags, I like the "ah, just drop it completely" idea here.

I don't like the mangling between feeds and URLs. We should get a good feedview, and if you add tags there, fine. If you add tags to a page, it should just be the page. In my experience, feeds and the webpages linking to them are not really equivalent that often. --AxelHecht 10:30, 27 Oct 2005 (PDT)

Bookmarking and Tagging

Bookmarking has definitely changed with the evolution of the web (http://pubnotes.wordpress.com/2007/10/14/evolution-of-bookmarking/). It is now time to rethink about the UI to address the new requirements. The current interface for BOOKMARKS / FAVORITES is optimal for "quick access to a small number of items with a high frequency of revisiting". But the "large number of items with a low frequency of revisiting" needs a new concept (http://pubnotes.wordpress.com/2007/10/14/rethinking-bookmarks-ui/). I consider it an INDEX / PERMANENT HISTORY of websites. Just saving these websites as bookmarks doesn't result in anything useful because organizing a directory or scanning the large list manually takes to long, and the saved titles usually don't contain the search terms I use. However, annotating the titles of bookmarks has been very useful before tagging became an issue (see above "Evolution of Bookmarking"). It turned out that commas between tags didn't add anything. Instead of tags with multiple associated words I use multiple single-word tags separated by spaces as in natural sentences (also suggested by del.ios.us). This is sufficient for the search and allows me to use the concept of "ordered tags" (http://pubnotes.wordpress.com/2007/10/14/ordered-tags/).

Tagging UI

The tagging interface I like the best is del.icio.us' web interface for posting (opened and pre-filled with a bookmarklet). It shows all used tags as a tag cloud and supports find-ahead suggestions while typing. This has limited the tendency of folksonomies to create multiple words for one topic. And its has reduced random spelling errors and the parallel use of plural and singular. Putting tags in an one-dimensional list would result in an extremely long list with endless scrolling. This was acctually the reason I never switched in the past to an alternative bookmark managers with keywords. I see tags as a form of annotation. This semantic annotation depends primarily on my current interest. For that reason I would like to have an interface that allows me to see and change tags parallel to reading the website. May be in a Sidebar, "Bottombar", or in a dynamic tool like the find tool that is activated by a single key. But this tool should not be limited to tagging. It should also integrate a large field for free annotations. My search tools are currently not specific for tags but include everything: URL, title, tags, and annotations (http://pubnotes.wordpress.com/2007/10/14/firefox-search/). It fails only in a few cases when the search term is a very common word or pattern so that I get to many results. A suggest button, a filter that scans the content of the website for all my tags, an automatic language tag, and pre-filling by a manually defined URL -> tags list would be great. I guess the three-panel and two-panel concept would only work for "main tags" but not for common sizes of tag lists because of the access by scrolling issue. See for example http://johnvey.com/features/deliciousdirector/. Very useful would be a simple button that bookmarks the website and adds a predefined tag(s), e.g. 2read, without any other user interaction. And for power users, it would be a great advantage if the tool would support GET-like function as Flock does for some tools, e.g. chrome://browser/content/bookmarks/bookmarksManager.xul?url=http://www.foobar.org&tags=cool, so that bookmarklets can be used to develop quick solutions for specific problems.

Tagging + Ranking + Sharing gives very interesting possibility!

Tonight I've invented a great thing, but I felt that someone should have invented it before me :)

First I've found Places and that was 1/3 of my idea.

From there I surfed to this page and found next 1/3 of my idea.

Here is a remaining 1/3 part:

1) On History Integration: add a small "Rate" field permanently on the toolpane (listbox or slider). User can rate the current page by one mouse click and this info is stored in the history and shown on the next visit of this page - now user may change the rating. That's much easier than adding bookmark or tag, but allows to mark the page as interesting or important.

When user decides to tag the current page, then the rating should be set by default at least to 4 stars (of 5).

Tags/History search should sort results by rating.

2) If two people rate the same web pages similarly, then most probably they have similar interests, tastes, etc. So I ask to find in the central tag database "people like me", with similar "rating profile". We can calculate a coefficient of similarity of two people. Now for every page ever tagged we can calculate individual page rank as a sum over all ratings of the page multiplied by the coefficient of similarity of the user who rated the page and me. If the web page is interesting for "people like me", then most probably it is interesting for me. So the search results can be sorted by my individual ranks.

--YaP 17:46, 21 May 2006 (PDT)