folksonomy按照数学模型的理论推断的确能达到最终的见山是山,见水是水,见山不是山,见水不是水,见山还是山,见水还是水的这么一种由浅入深,有内到外,有最终形成一种提纲挈领的的判断.

为什么突然想起这个呢？今天看到ptree《同分妄见与家族相似》提到从哲学上，探讨本体论与分众论的历史渊源。很值得一读。其实期间还有值得挖掘的东西还很多
http://ecolab.ruc.edu.cn/blog/zhangsr.php?itemid=132

如果你已经看相关的关于Tags的各种说法，我希望你能够暂时忘掉全部的那些解释，之后再重新回忆起来；但如果你还不了解Tags，那么有必要先阅读文章最后留下的链接，然后回到上一步：）

　　那么，什么是Tags？很简单，往下看；

·经典的信息构成模式；

　　在传统的网页组成中，我们通常使用Taxonomy(分类法)来归纳、整理和存放我们的信息，图书馆是一个绝好的例子，所有的信息从一个点开始，形成树状的分类，由此构成一个完整的、相互联系起来逻辑体系；
　　这个体系从一开始就是人为分类形成的，在我们需要检索的时候，几乎不需要费多大的劲；

　　[例子]比如我们的Blog，首先有一个主标题，然后下分若干分类，实际的文章则分类储存在这些不同的分类中；在一般情况下，我们不允许一个文章同时存在于多个分类中，以便于我们管理的方便和检索的唯一性；

　　在网络上,dmoz和wiki都算是比较典型和知名的例子；

·散秩的信息构成模式；

　　看似无序的浩渺信息，其中的绝大部分是通过“语言”来描述的，这表明了这些信息的指向性，因此我们通过提取这些语言（文字、文本）的相同部分，以此获得相关的信息；这些信息平时是完全松散、互不联系的，当且仅当我们对其加以提取的时候才呈现相对紧凑的组织结构，即便如此，这个结构和经典的分类结构相比，仍然是足够散秩的。

　　[例子]你可能已经想到了，Google嘛。目前绝大部分的搜索引擎所依赖的正是这一点，因此对于分词的研究是这些搜索引擎始终的重点和痛苦，其他的不说，仅仅逻辑实证主义和日常语言学派这两个当代的流派就足够他们折腾到下个世纪去了。
　　举一个恶搞点的例子：当我说：“他妈”的时候，仅仅检索关键字而并不关心其在日常语言中实际运用的搜索引擎怎么知道我是在骂人还是在陈诉一个归属性事实呢？更何况我们经常面对google上数万和关键字原本语义要求完全不同的搜索结果长声叹息。

　　[简介]
　　逻辑实证主义：认为人类的日常语言充满的谬误，需要彻底厘平，重构一个像数学一样完美的逻辑语言体系；
　　日常语言学派：认为人类的日常语言是非常合理而符合现实的，“完美”的逻辑语言并不存在而且也不符合现实；唯一的问题在于人们使用日常语言的时候出了一些方法上的问题，这需要我们加以重视和研究。
　　（后者正是我倾向赞同的结论）

·符合未来发展的信息构成模式；

　　现在我们综合起来考察以上两种在我们日常生活中显得日益重要的信息构成模式，会发现他们各有优点和缺陷；
　　对于前者而言，语言所表达和内涵的思想是广博的，构造简单的分类逻辑无法诠释和标识某一篇文章所设计的全部重点，复杂的分类则将陷入无限微观的悖论逻辑；
　　对于后者而言，除开分词的烦恼，Google们也许还希望承天下之大义担负起教导每一个人重修日常语言学分的重任，并且要求每一个人都能达到维特根斯坦的高度。

　　路德维奇·维特根斯坦？对了，这终于回到我们的重点。
　　维特根斯坦本人正是逻辑实证主义和日常语言学派先后的奠基人，而在他的后期的日常语言思想中，他提出了一个大家相对比较熟悉的观点：家族相似。

　　以下引用一段话作大致的解释：

    维特根施坦从“反本质主义”立场出发反对语词的定义化。本质主义者认为同一类事物之所以成为该类事物，是由于它们具有共同的本质（共相），定义就是规定事物的这种本质。维特根施坦则认为事物根本没有共同的本质，只有“家族相似”。所谓“家族相似”不是共同的相似，而是这一方面或那一方面的不完全相似。例如一个家族中的成员之间有的眼睛相似，有的神态相似，有的脸庞相似。因此，维特根施坦坚持一种唯名论的立场，认为人们在日常生活中使用一般性的名词概念只是为了方便，本质、共相那种形而上学的东西是不存在的。误把这些东西当作存在，就会染上“哲学病”。

2005-3-27 17:15:51  Aether answer vistor:我觉得有必要讨论tags应用的两个重点，
　　1、适用范围；
　　Tags应用必然是针对广泛、相对散秩同时有更高质量要求的信息群落；（和生命本身一样，也是整体无序中产生的部分有序）
　　这个无序的群落要构成足以产生有序的宏观规模，少则几千，多则应该上万，数十万；但是同时不能无限制膨胀，规模再高，Tags的作用可能就不及完全无序的关键字了；
　　因此仅仅在一个个人Blog的内部，为区区数十上百条信息建构相对散秩的Tags，实际意义不大，因为对于少量信息人为的分类应该已经足够；而通过统一的标准加以组织、集中和聚合，才有可能形成有效的规模；

　　2、关键效用；
　　Tags产生以后，在检索的过程中，和关键字检索几乎没有任何区别；因此Tags的关键部分在于它产生的过程：广大的用户群对录入的信息人为地加以预筛选。
　　因此这要求对录入的过程有比较有效的监督；同Wiki一样，这种开放式的预筛选过程不同于分类法的完全封闭和控制，这方面的研究还有待深入和继续。 (via here)

The past few weeks have seen my inbox flooded with folksonomy questions. I am going to make things easier on my inbox by posting some common discussions here. Many of the items I am posting I have posted else where, but this will also be a great help for me.

There have been many people who have correctly discerned a difference between the two prime folksonomy examples, Flickr and del.icio.us. As I first stated in a comment to Clay Shirky's first article on Folksonomy, there are two derivations of folksonomy. There is a narrow folksonomy and a broad folksonomy. On August 26th I stated...

Clay, you bring in some very good points, particularly with the semantic differences of the terms film, movie, and cinema, which defy normalization. A broad folksonomy, like del.icio.us, allows for many layers of tagging. These many layers develop patterns of consistency (whether they are right or wrong in a professional's view is another matter, but that is what "the people" are calling things). These patterns eventually develop quasi power law for around the folk understanding of the terms as they relate to items.

Combining the power tags of "skateboarding, tricks, movie " (as you point out) will get to the desired information. The hard work of building a hierarchy is not truly essential, but a good tool that provides ease of use to tie the semantic tags is increasingly essential. This is a nascent example of a semantic web. What is really nice is the ability to use not only the power tags, but also the meta-noise (the tags that are not dominant, but add semantic understanding within a community). In the skateboarding example a meta-noise tag could be gnarly that has resonance in the skate community and adds another layer of refinement for them.

The narrow-folksonomy, where one or few users supply the tags for information, such as Flickr, does not supply power tags as easily. One or few people tagging one relatively narrowly distributed item makes normalizing more difficult to employ an tool that aggregates terms. This situation seems to require a tool up front that prompts the individuals creating the tags to add other, possibly, related tags to enhance the findability of the item. This could be a tool that pops up as the user is entering their tags that asks, "I see you entered mac do you want to add fruit, computer, artist, raincoat, macintosh, apple, friend, designer, hamburger, cosmetics, retail, daddy tag(s)?"

This same distinction is brought up on IAWiki' Folksonomy entry.

Since this time Flickr has added the ability for friends and family (and possibly contacts) to add tags, which gives Flickr a broader folksonomy. But, the focus point is still one object that is being tagged, where as del.icio.us has many people tagging one object. The broad-folksonomy is where much of the social benefit can be derived as synonyms and cross-discipline and cross-cultural vocabularies can be discovered. Flickr has an advantage in providing the individual the means to tag objects, which makes it easier for the object to get found.

This brings to the forefront the questions about Google's Gmail, which allows one person the ability to freely tag their e-mail entries. Is Gmail using a folksonomy? Since Gmail was included in the grouping of on-line tools that were in the discussion of what to call these things (along with Flickr and del.icio.us) when folksonomy was coined I say yes. But, my belief that Gmail uses a folksonomy (regular people's categorization through tagging) relates to it using the same means of one person adding tags so that object can be found by them. This is identical to how people tag in Flickr (as proven by the self-referential "me" that is ever prevalent) and del.icio.us. People tag in their own vocabulary for their own retrieval, but they also will tag for social context as well, such as Flickr's "MacWorld" tags. In this case Wikipedia is a little wrong and needs improving.

I suppose Gmail would be a personal folksonomy to the Flickr narrow folksonomy and the del.icio.us broad folksonomy. There are distinct futures for all three folkonomies to grow. Gmail is just the beginning of personal tagging of digital objects (and physical objects tagged with digital information). Lou Rosenfeld hit the nail on the head when he stated, "I'm not certain that the product of folksonomy development will have much long term value on their own, I'll bet dollars to donuts that the process of introducing a broader public to the act of developing and applying metadata will be incredibly invaluable.". These tools, including Gmail, are training for understanding metadata. People will learn new skills if they have a perceived greater value (this is why millions of people learned Palm's Graffiti as they found a benefit in learning the script).

Everybody has immense trouble finding information in their hierarchal folders on their hard drive. Documents and digital objects have more than one meaning than the one folder/directory, in which they reside. Sure there are short cuts, but tracking down and maintaining shortcuts is insanely awkward. Tags will be the step to the next generation of personal information managment.(via here)

Last week I asked the AIfIA members' list what they thought about the social classification happening at Furl, Flickr and Del.icio.us. In each of these systems people classify their pictures/bookmarks/web pages with tags (e.g. wedding), and then the most popular tags float to the top (e.g. Flickr's tags or Del.icio.us on the right).

Thomas Vander Wal, in his reply, coined a great name for these informal social categories: a folksonomy.

I think folksonomies can work well for certain kinds of information because they offer a small reward for using one of the popular categories (such as your photo appearing on a popular page). People who enjoy the social aspects of the system will gravitate to popular categories while still having the freedom to keep their own lists of tags.

On the other hand, I can see a few reasons why a folksonomy would be less than ideal in a lot of cases:

• None of the current implementations have synonym control (e.g. "selfportrait" and "me" are distinct Flickr tags, as are "mac" and "macintosh" on Del.icio.us).
• Also, there's a certain lack of precision involved in using simple one-word tags--like which Lance are we talking about? (Though this is great for discovery, e.g. hot or Edmonton)
• And, of course, there's no heirarchy and the content types (bookmarks, photos) are fairly simple.

Still, the idea of socially constructed classification schemes (with no input from an information architect) is interesting. Maybe one of these services will manage to build a social thesaurus.(viahere)

Dave Weinberger talks about Folksonomies (tagging photos, web pages etc.), and the problem of how tags aren’t controlled; thus, the one person may use the tag “beach” while the other may use “shore.”

I could imagine one simple solution to this: let the same collaborative filter that works for tagging be applied to finding synonyms. There must only be a mechanism in tagging software which allows people to say “beach is synonymous to shore”. This is not a clear true-or-false situation, which is why one would take the percentage of people saying “A is a synonym of B” to determine the synonym strength. This strength could then be used to adjust the system’s output.

Let’s give an example. I subscribe to various RSS feeds serving content for the tag “google”. If a majority of taggers would now decide that “googling” is 70% of “google”, and “search” is 20% of “google”, then in my RSS feed I would now also get a random 7 of 10 items which were tagged with “googling”, and 2 of 10 items which were tagged “search”.

Because people may strengthen synonym relationships of popular words out of proportion, one might need to be able to vote for antonyms as well to create a balance. In semi-random intervals, one would need to ask taggers if two randomly selected words are related or not. This may even be an application for CHI...(via here)

关于Folksonomy（分众分类）与语言的关系是一个很有意思的问题。

Tag的兴起源于两种社会性网络服务的应用 - del.icio.us书签和Flickr的图片分享服务。这两种Web服务很巧妙的将Tag功能运用其中，配合他们独特的易用性，一定程度上引爆了Tag以及分众分类的流行。不过这些服务的语言系统都是基于英文，所以在接受和认知程度上，他们更符合英文用户的使用习惯。

分词特征：英语和汉语在分词的定义上有着显著的区别。英文中的词干以单词形式出现，而中文中的词干是由单个字组成的“词组”。从语言学（Linguistic）的角度来说，英语中的单词可以直接排它性的定义，而汉语的词需要通过特定的分词结构来进行区别。举一个简单的例子，“她用魔法做了一个毒苹果然後扮成一个老婆婆到森林去找白雪公主”。这个句子中的“苹果”和“果然”就需要通过中文的分词算法来解决，到底是“苹果”为一个词还是“果然”为一个词？而英语中很少会出现这样的问题。总的来说，两种语言的分词特征有显著不同。关于中文分词详情见卢亮的中文分词系统。

语义普遍认知（Common Sense）的差别：在不同的语言环境下，对语义的普遍认知存在很大的差异，比如Technology，这是一个普同的定义词，但在汉语中，对于它的解释可能是：科技，技术，工艺等等。在这种情况下，Technology本身在英语语境的普遍认知度则比科技，技术，工艺这三个词中的任一词在汉语语境中的普遍认知度高。同样的情况也出现在汉语语境中，Keso举了“春节”这个例子，在英文中它可以表述为ChineseNewYear，SpringFestival，LunarNewYear等等。可见，不同语言系统中的语义普遍认知度差别明显。

缩写词的特征：这是在英语中一个很特有的现象，比如Get Things Done这个短语是一本书名，在使用中往往被缩写为“GTD”，又比如中国共产党 - Chinese Communist Party，在使用中被普遍缩写为CCP。在英语语境中，许多特定词组都是通过缩写形式来表达。而这种表达方式运用于Tag标签就非常方便，如创作共用Creative commons往往被缩写为CC, 这样的例子不胜枚举。但在汉语语境中则没有这个特征。

趋同性：在整个Tag系统里面还有一个很显著的特征，就是趋同性。也就是对于同一个对象的定义，往往人们趋同选择被定义次数多标签作为其分众分类。举个例子，boingboing在美味书签中被使用“Blog”标签了200多次，而被使用“news”标签了90次，也就是说，用户在定义的时候更趋向于选择使用次数多的标签。而这种结果又恰恰的反射出大家对同一事物的普遍认知度上。所以在整个Tag系统中，用户对一个事物的理解有着趋同性的特征。并且这种趋同性从一个侧面能够反映出群众的普遍认知度，甚至用户结构等。

通过最近做的一些简单使用调查，结合开发中的一些经验，对于分众分类在在汉语上的应用和普及我持比较保守的态度。不伦是从语言的使用习惯，用户群的普遍认知度，还是汉语语境的分词结构，许多地方分众分类标签容易模糊用户的导向，分众分类在中文的社会性网络服务中作为一种辅助功能而存在更为可行，或者在某一些特定领域比如具体地点名称上进行应用，但作为内容平级分类模式，其应用与汉语语境的用法和习惯等诸多方面无法协调，它并不是一种能够在中文环境中保持生命力的应用模式。

I was sitting up and got pinged by Dave Sifry about Technorati’s new related-tags feature; Technorati thinks that Baseball is related to Sports, MLB, Football, Basketball, Natural Philosophy (gotta love that), and tickets. Some don’t work that well, but the idea is compelling. I’ve been thinking about this stuff a lot, and I have a question: Do tags work? It shouldn’t be too hard to find out.

Someone needs to do a little basic research. First (this is the hard part) you pull together a set of sample searches. You could use a recent searches page, or I bet Dave Sifry would cough up a random selection from the Technorati feed.

Then, you go and do each search twice, once using keywords and once using tags. Note that the two searches can be different, it’s OK to apply human judgment. Then you record which way worked better (once again, there’s some subjectivity here). Then you write up your results.

I don’t have time to do this research, but if it were done reasonably transparently and carefully, I’d certainly contribute a pointer, and since quite a few people who care about search read ongoing, the word would get around pretty quick.

folksonomy club: Folksonomy: social classification: "I think folksonomies can work well for certain kinds of information
because they offer a small reward for using one of the popular
categories (such as your photo appearing on a popular page). People who
enjoy the social aspects of the system will gravitate to popular
categories while still having the freedom to keep their own lists of
tags."

Synonym control is a real concern, and I think you'll start to see tools emerge on popular folksonomies as they become bigger and bettter.

I'm not sure about the problems as content types get more complex, though it is something to think about.

The other two concerns are, I think, are either solvable or not a big problem. For instance, using cluster analysis tools, you can figure out what words tend to match up with others. Going one step forward with that, you'll probablly be able to find out what word clusters are seperate from each other (for instance if I have Lance Armstrong photos, I'll probablly also add bike or tourdefrance or something like that, but if I've got pictures of medieval lances, I'll add weapon or knight. Since the words that cluster with lance don't match, I can figure out if there's a homonym divide by providing a second level cluster analysis and then advise a user that they might like to add a distinguishing tag at both creation time and search time) or, the thing that is being referred to will be so general, that they won't be in the shared taxonomy and thus would not need to be disambiguated from the rest (If you had pictures of your cousin Lance and I had ones of my brother Lance, there wouldn't be a need for disambiguation since the seperate Lances each reside in our own personal taxonomies and don't get dragged into the folksonomy very much).

The lack of hierarchy, on the other hand, I think is the STRENGTH of the folksonomy and I see a lot of people raising this point. It seems like they're almost afraid of hierarchies going away because they're so familiar and comfortable with them. In a folksonomic system, you can just be rid of folksonomies because you can rapidly BUILD whatever hierarchical category you want by combining tags. If I want to find things about apples that are fruit, I would add apple and fruit together in my tag search. If I want further than that, Granny Smith apples, I'd just throw grannysmith on there. On the other hand, suppose I just want to see grannysmith right from the start. Imagine I needed to navigate a sub directory to do that? Well, first I'd have to figure out which top level directory it was in, then the next level and the next. While I'm pretty sure that no one's outright suggesting that we start building big hierarchies around these things, I think that once you start doing that, you find it hard to stop because as you get more and more information, you need more and more ways to classify it. Or, alternatively, you decide to omit certain kinds of information because they would force the hierarchy to get bigger. Either choice is unappealing.

The reality is, and I'm perfectly ready to acknowledge this, that you lose information in a folksonomy. The difference here is that people are willing to both accept this loss and participate because of the advantage gained, both personally and globally. Compare that to a formal classification scheme where people aren't willing to participate. Sure you might (and I think this is an important point, because outside of laboratory conditions, its hard to understand everything that defines things that need classification) capture more information, but you're going to get less and, possibly, people aren't going to care enough to use it.(via here)

Wists help中的这段话帮我更深刻的认识了folksonomic：

     Advanced tagging: Wists allows you to create groups of tags, called 'themes', in order to manage large numbers of tags better. For example: to bookmark a Sushi Restaurant in New York you could enter: Restaurant location=ny type=sushi. You can invent as many themes as you like.) Multi-word tags: use underscores if your tag names are phrases e.g. latin_america. （Via here）

其实传达了人们使用分众分类要求的自由和无拘束。分众分类带着浓厚的blog色彩。你对他进行分类和摘要的时候已经表达了你的立场、观点。真正实现用我们自己的想法区分这个世界(Classify the whole world by ourself)!（Via here）

Last week I asked the AIfIA members' list what they thought about the social classification happening at Furl, Flickr and Del.icio.us. In each of these systems people classify their pictures/bookmarks/web pages with tags (e.g. wedding), and then the most popular tags float to the top (e.g. Flickr's tags or Del.icio.us on the right).

Thomas Vander Wal, in his reply, coined a great name for these informal social categories: a folksonomy.

I think folksonomies can work well for certain kinds of information because they offer a small reward for using one of the popular categories (such as your photo appearing on a popular page). People who enjoy the social aspects of the system will gravitate to popular categories while still having the freedom to keep their own lists of tags.

On the other hand, I can see a few reasons why a folksonomy would be less than ideal in a lot of cases:

• None of the current implementations have synonym control (e.g. "selfportrait" and "me" are distinct Flickr tags, as are "mac" and "macintosh" on Del.icio.us).
• Also, there's a certain lack of precision involved in using simple one-word tags--like which Lance are we talking about? (Though this is great for discovery, e.g. hot or Edmonton)
• And, of course, there's no heirarchy and the content types (bookmarks, photos) are fairly simple.

Still, the idea of socially constructed classification schemes (with no input from an information architect) is interesting. Maybe one of these services will manage to build a social thesaurus.(Via here)