*****  To join INSNA, visit http://www.insna.org  *****

Here is a collection of answers that I got so far plus some of my ideas 
for solving the problem. I would like to thank everybody that responded 
and urge everyone to discuss the proposed solution.

Let me first repeat what I'm interested in. I am especially interested 
in how do you create a network (define ties) from online forums where 
people can not reply (they can however cite) to individual posts, but 
just add new posts to the thread/theme (or you can say reply to the theme).

Mostly, I would like to thank Paul, who has faced exactly the same 
problem. He said he found no way to automate the creation of 100% valid 
ties. Her tried 3 naive approaches:
1. A post has a "To" tie (in the e-mail sense) to the post post directly 
in front of it and a "Cc" ties to all posts in front of it.
2. A post has a "To" tie (in the e-mail sense) to the post post directly 
in front of it.
3. Construction of the two-mode network (posts and threads) (the same 
suggestion was made by Adam).

He also pointed out the advantages and disadvantages of all 3 options. 
My idea is similar to his 1. approach. The disadvantage of this approach 
is that later posts might have very little to do with some posts that 
were posted quite early in the discussion. My idea is to first limit the 
previous posts to which a new post creates a "cc or to" tie by the 
number of posts between them and the time elapsed between the two posts. 
I also intend to weight (negativly) the tie by the distance in "queue" 
and time. Do you think that that would be a good representation of the 
distance between the two "posts" and their posters.

The third approach is approach via two-mode networks. This idea is 
nicely explained by Adam who wrote:

    I considered working on a project with online forums, but it ended
    up falling through, so I have no experience per se and don't have a
    paper to send you. But my thought at the time was to consider the
    online forum network to be a two-mode network of people and forums. 
    Thus, a tie is defined as co-participation in the same forum.

    Needless to say, you'd like to add in layers of complexity that will
    probably depend on both your research question and your data.  For
    example, you could define tie strength as some function of the
    number of submissions two people make to the same forum.  If two
    people co-submit to multiple forums, that could be an indication
    that they are extremely similar, so you could give that a
    multiplicative effect rather than an additive effect.  You could use
    time-limiters (assuming you have that data), such as (1)
    co-submission only results in a tie if your submission happens
    within a month of mine or (2) my "membership" in the forum lasts
    from my first post to my last, so I'm only tied to anyone who posts
    during that interval.  Etc.


A would like also thank Alvin who described his experience with blogs 
and newsgroups. Although blogs and newsgroups are not so problematic in 
terms of defining links, I found his response usefull nevertheless, and 
it also includes a link to a several of his papers on this topic:

    I'm actually doing this myself with blogs and newsgroups.  My
    experience is that I create the social network where each link is from
    the author's post to the author's comment to that post, and also the
    author's comment to the previous comment since there can be a possible
    social tie between commenters or between the commenter and the author
    of the post.

    Hope that helps, and if you want more info, you can see some of my
    papers that I've published on this at
    http://www.imedia.mie.utoronto.ca/~achin/publications.htm. Look at
    referred conference papers.

Thanks to all,
Ales


P.S.: I hope that all who have emailed me privately do not mind sharing 
their response on the list, as nobody indicated so. Answers from some of 
the responses are not included her, as I am still concussing their 
responses with them and will post the final conclusions were appropriate.




> -----Original Message-----
> From: Social Networks Discussion Forum [mailto:[log in to unmask]] On
> Behalf Of Ales Ziberna
> Sent: den 21 maj 2007 21:53
> To: [log in to unmask]
> Subject: Extracting (social) networks from online forums
>
> *****  To join INSNA, visit http://www.insna.org  *****
>
> Dear list,
>
> I'm looking for any experience with creating (social) networks from 
> online forums. Any material on that subject would be most welcome.
>
> The main problem is that the definition of a tie is quite problematic in
>
> such cases, as the forum in question does not allow people to reply to 
> posts, just add posts to a thread. They can cite posts, but this happens
>
> quite rarely.
>
> So if anyone have any experience whit this, I would be very grateful for
>
> you help.
>
> Best regards,
> Ales Ziberna
>
> _____________________________________________________________________
> SOCNET is a service of INSNA, the professional association for social
> network researchers (http://www.insna.org). To unsubscribe, send
> an email message to [log in to unmask] containing the line
> UNSUBSCRIBE SOCNET in the body of the message.
>   

_____________________________________________________________________
SOCNET is a service of INSNA, the professional association for social
network researchers (http://www.insna.org). To unsubscribe, send
an email message to [log in to unmask] containing the line
UNSUBSCRIBE SOCNET in the body of the message.