People-marking: People-Centric Bookmarking

del.icio.us and other social bookmarking tools are just the beginning of a shift in the fundamental browsing process. They have shown that, if tools make it easy enough, users will take the time to categorize and annotate as they browse, and in doing so create a double-filtered web that increases the productivity of all. Many web citizens will be come true researchers– readers and writers, and will share the fruits of their research as never before.

The technology for this collaborative research will evolve from the document-centric and textual tagging basis that spawned it. People, now treated as second-class subjects in relation to documents, will become a central focus. This transformation has begun with the social networking revolution, but thus far has remained within the domain of dating and job hunting, and within the “this-is-me” and “these-are-my-friends” framework.

Persons-as-subject is the next step in the evolution. We all google, not just to find documents, but to find people! But a person is a nebulous object on the web, akin to Sybil, the famous woman with nineteen distinct personalities. Each of us has a scattered presence on the web, but current research tools like del.icio.us don’t provide much help in allowing users to create an organized virtual person.

With del.icio.us, you can tag a document with a person’s name, e.g., ‘david_wolber’. Readers can then go look at all the documents under ‘david_wolber’. But ‘david_wolber’ is just text. You can’t set a homepage for the person David Wolber, or specify that he has a blog and a del.icio.us user name. And you can’t do things with David Wolber, like put him in a group, or associate him with another person.

People-centric bookmarking improves upon the process by making people first-class objects and giving them at least equal emphasis with documents. Users can create person objects, not just for themselves or their friends, but for anyone whom they are researching. Users can then ‘grab’ one of these virtual people and do things with him– associate the person with a document, i.e., people-mark the document, add the person to groups, and associate the person with another person. Search in such a system returns both people and documents.

peoplicious.com is a people-centric bookmarking site. Check it out!

Social bookmarking in corporations

Mitre research has joined IBM as a company exploring intranet bookmark  systems for their corporations. Mitre’s is called Onomi, while IBM’s is called dogear.

I’m listening to Laurie Damianos of Mitre speak at the HICSS conference. One of their goals is expert finding, and Onomi does allow nice profiles for users, which enhances expert finding within the company. They don’t have a concept of profiles for non-users.

Wikipedia vs. Peoplicious

I’m trying to figure out the essence of peoplicious and specifically how it fits into the world which has something called wikipedia. The short answer is that peoplicious is about entering structured information and links using forms and bookmarklets, while wikipedia is about entering biographies about people using html and wiki format. Peoplicious provides a less-effort collection bin of info and links about a person. Inputting articles about a person is integrated with the browsing process– more like bookmarking than wiki-ing. Wikipedia provides a more-effort method of compiling prose and links about a person.

Their goals are different as well– Peoplicious is  about collecting all the stuff about a person on the Internet, including recent news, while Wikipedia is about compiling information that would go into a person’s permanent biography. For instance, peoplicious structured data includes name, image url, homepage url, blog url, del.icio.us url, and technorati name. Wikipedia structured data for a person consists of the fields:

<!-- Metadata: see [[Wikipedia:Persondata]] -->
{{Persondata
|NAME=Magellan, Ferdinand
|ALTERNATIVE NAMES=Magalhães, Fernão de (Portuguese); Magallanes, Fernando de (Spanish)
|SHORT DESCRIPTION=Sea explorer
|DATE OF BIRTH=Spring [[1480]]
|PLACE OF BIRTH=[[Sabrosa]], [[Portugal]]
|DATE OF DEATH=[[April 27]], [[1521]]
|PLACE OF DEATH=[[Mactan Island]], [[Cebu]], [[Philippines]]
}}

Input Process
Wikipedia allows users to enter information about a person in the form of HTML wiki pages. Guidelines are provided to help standardize the look of these pages.There are also ‘structured’ elements. One can put a person’s page in various categories by providing a link on the page to the category page. Users can also add PersonData, which is a set of biographical fields meant to be read by other software and which is by default hidden from wikipedia users.

To add a person in wikipedia, a user must learn html or the wiki formatting. This is not too complicated, but is enough to lose a lot of potential contributors. They must also learn the guidelines set-up for biographical pages. This requires at least a few minutes of research.

Peoplicious is form-based, so no html or wiki protocol is required. The user clicks on contribute person, then enters the person’s basic data in the dialog. Wikipedia allows the user to enter prose about a person as well as links. Peoplicious is not for adding prose– its for adding links about a person. Even the images for a person are just links.

Peoplicious also provides bookmarklets and a firefox plug-in that makes adding an article to a person’s page extremely easy and integrated with the browsing process. This is called people-marking an article. With wikipedia, there is no such concept– all information including links are just html, so the user must go to the page for the person and edit it. There is not a huge difference in effort, but one must always keep in mind that browser-researchers have a very low effort-threshhold when it comes to pausing their research to record their findings.

Adding people to lists in peoplicious is also form-based– the user selects the person from one listbox and the list from another listbox. With wikipedia, the user enters a link to the ‘category’ page using the wiki format for an internal page.

Wikipedia Documentation
Information about categories. You add a page to a category by inserting a link in the page to the category page.

Information about People Categories

PersonData is metadata for biographies. Germans are using it a lot. No xml interface to it yet.

Project Biography is a working group on wikipedia dealing with bios.

Tagged!

tagged is a teen social networking site that allows people to be tagged.

Jigsaw’s shared rolodex

jigsaw is something like Plaxo in that its an on-line address books. It tracks companies, individuals, etc. and gives people a way to share their business contacts. Esther Dyson says it allows people to sell out their rolodex. She describes its essence this way:

The concept of Jigsaw is this: You take the contacts you have, and put them online. For each one you post, you get credits that you can use to buy other contacts. Or, you can simply open an account to buy contacts at $1 each.

Should lists replace tags?

What is the difference between a list and a tag?

(del.icio.us) tags are restricted to not having spaces, of course. Part of the reason for this is to make it easier to enter tags (don’t have to type commas on the tagline if you want to enter more than one tag for an item).

We might say that lists, on the other hand, can have a name with multiple words. For some systems, this might slow down the tagging process.

We might also say that a list is more than just text,– it can have data associated with it. For instance, a list representing an organization might have that organization’s home page.

John Tropea, in discussing lists on the web, points out how tags don’t allow this:

I suppose you could even use social bookmarks like del.icio.us, the problem here is that the tag label is the heading, leaving you nothing to describe the list with.

Basically, with a tag, there is no thing to describe or to tag. With a list there is.

With peoplicious, you don’t tag people, you put them in lists. Lists can have data associated with them. I don’t allow the lists themselves to be tagged, as John suggests.

With peoplicious, you people tag articles– you bookmark the article and say that it is about or by a person in the system. Those articles then appear under the person and whatever lists the person is in.

After reading John’s post, I was reminded of an earlier idea I had which was to also allow articles to be bookmarked directly under a list, and not one person. For instance, in the political peoplicious, http://opencampaigns.com, there is a list for SF Supervisor District 6 candidates. For an article talking about all the candidates, one has to categorize it under each individual in that list. It would be nice to just say its about the whole list.

In some sense, this would allow tagging (listing) an article by topic, where currently you can only people tag. Hmmm….

Wordcamp list

I’ve set up a peoplicious list for wordcamp presenters, its at http://peoplicious.com/Technorati/lists/Wordcamp%2006.

Feel free to add to it…doesn’t have to be just presenters.

peoplicious is just a prototype so may blow up if many people look at it.

People Search

Here’s the problem: I’ve got a system (peoplicious) that allows users to collaboratively create virtual people. A virtual person is defined by some profile info (name, etc) and a list of urls including homepage, blog, blog posts, delicious posts, and personmarks (documents that the person or others have marked as being authored by/about that person). A person can also be tagged directly, so that set of tags are thrown in the mix (as well as the tags the person uses on his posts).

So the system has these virtual people, and you want users to be able to enter keyword searches that return not documents but people, e.g., experts on a topic. So the question is: how do you build your search index?

With regular document search, you put all the keywords of a document into the index, and when those keywords are searched for, that document is found (and listed if its rank is high enough).

Now in indexing virtual people, we have to decide:

1. What documents should be indexed and how it should be weighted (e.g., should homepage text be more important than blog post text)

2. For each docment indexed, what terms should be included. Because we have n documents for a person, instead of just one as with document indexing, we may want to be more selective in the terms that are added to the index, e.g. if ‘folksonomy’ if found only once across all documents defining a person, that person is probably not an expert.

One solution, I guess, is to come up with a frequency list of all terms with the person’s defining documents. That is, don’t index each document, but first build a frequency list of terms in all documents, then index the top terms (using tfidf with probably a lot of weight on the term frequency).

Currently, peoplicious just takes a person’s homepage and blog and throws all the terms in an index, and that’s search. Users also see peopletags (called lists) on the main page and can just choose the list to see persons with that tag.

I haven’t mentioned ranking– right now there is none, but there should be.

Peoplicious: The People of Google

I’m visiting Google to talk about peoplicious next week, so I thought I’d set up a peoplicious site for The People of Google. Right now only I, the developer of peoplicious, can create a new domain. I do this by modifying a couple database tables and it takes a couple of minutes.

One thing I’m contemplating is allowing domains to be created by end-users. By domain, I mean a separate site, a separate people directory, where all the lists are subordinate to, e.g., The People of Google. Kind of like PeopleAggregator allows for social networks. Prior to setting up the Google site, I set up a domain for techies in general, for my university, USF, and for politicos.

Google, as an example of a corporation with highly skilled employees, is an interesting case. They have some incentive to not have their people well known (Microsoft might poach them!!!). Peoplicious, unlike most social networking tools, allows a user to create virtual people other than herself, and tag those people (put them in lists). So it could be used for a research (team) to create a map of google employees. Because its not hierarchical and you can put people in various lists, it could show all the formal and informal communities in the organization. Not sure if this is something a Google would want.

I’m still working out how domains will work (peoplicious is a working prototype). The way it works now, people are in a single general table for the whole world, and can be part of various domains. Each domain has its own lists. There are no links between domains, so one believes they’re at a completely separate site (and domain names could be setup for this). One question is: when someone peoplemarks a document to a person, should that document appear in all domains for that person. If the answer is no, we in some sense are setting up personalities, i.e. split people…but there needs to be some way to integrate a virtual person.

SYBIL: Multiple Personalities on the Web

Denis McDonald has an excellent comment on peoplicious and other aggregators/expert systems, concerning the fact that a person/blog is often about many topics, and most tools don’t allow for multiple personalities. Here’s an excerpt:

Weaknesses include possible deficiencies in tagging the content of individual entries that are read into Peoplicious via the RSS feed that supplies the source blog’s information. For example, all my recent blog postings are included in the entries under “expert systems,” even those that have nothing to do with expert location or expertise management.

The flip side of the above is that the reader is exposed to things he or she might not otherwise be exposed to; making “people tagging” the basis for the grouping of items is a valuable approach.

Blog input tools provide categories to help organize one’s multiple personalities. However, aggregators, at least peoplicious, only let a user tag an entire person into a list, and there is no way to specify that a person has multiple areas of interest.

So we might think of allowing a single person to be subdivided, different feeds be set for the different personalities (given their blog input tool allows for different feeds for different categories), and then allow personalities instead of full persons to be added to lists (tagged). Then I could put Dennis McDonald:Expert Systems into  the Expert Systems folder.

Would this be too complicated? Are people not ready to admit multiple personality disorder? If only Freud were alive…