Twitter import (search) suggestion

Nov 15, 2009 at 4:10 PM


Would it be possible to add an option to the Twitter search import tool which linked the Tweeter with *any* @ that was mentioned in their Tweet?  For example in this example Tweet: -

Tweeter: RT @Person1: RT @Person2: Respect to @Person3 - did her talk at #hashtag with her new baby (pic via @Person4)
The resulting pairs would look like this: -
This would allow us to capture mentions, replies to and re-tweets and would be very, very useful.
Kind regards, hudmb.
Nov 15, 2009 at 9:54 PM

If I understand your suggestion, the Twitter search network, which is currently a network of people whose tweets contain a specified search term, would become a network of people whose tweets contain a specified search term AND the people mentioned in those tweets, even if the people mentioned didn't tweet the specified search term.  I find that awfully confusing.

Perhaps this could be an additional type of network, but I don't think it fits into the current network.

-- Tony

Nov 16, 2009 at 10:22 AM
Edited Nov 16, 2009 at 10:26 AM

Hi Tony, maybe I didn't explain it properly. 

What I'm after is the association between the tweeter and the people mentioned in their tweets, regardless of whether it's just a mention, or a reply or a re-tweet, when a certain search term is present.  So, say if I was to search for NodeXL on Twitter: -

There are people who refer to other people in their tweets.  Just to pick a few examples with what I would like to see as pairs extracted beneath: -

kremplo: RT: @nodexl: NodeXL v.98 ships, now with YouTube user network import #SNA #socialmedia - visualize vlogger connections


qualintitative: RT @marc_smith NodeXL v.98 ships, now with YouTube user network import #SNA #socialmedia


SameerPatel: Go Marc! RT @marc_smith: NodeXL v.98 ships, now w/ YouTube user network import > visualize vlogger connections


raquelrecuero: RT @marc_smith: NodeXL v.98 ships, now with YouTube user network import #SNA #socialmedia - visualize vlogger conne ...


Kueynislan: Para quem trabalha com análise de redes sociais, o @marc_smith anunciou que o @nodexl (software para ARS) agora tem twitter.


seekoeur: @sirchamallow @Alex_Riopel @Leslie_ @netintelligenz @dpillaert @adaptateur @CaddeReputation @gilldelia @VeroniqueR @markich @nitramf #NodeXL


So I'd like a list of pairs built up between the @tweeter and those @mentioned in the tweet for a given search term (in this case "nodexl").

I did this manually for a conference I went to where a specific #hashtag was used.  I searched for the hashtag on Twitter and then manually created pairs (as above) to figure out who was the most "influential" Twitterer at the conference based on who they referred to in their tweets (and of course, who referred to them in their tweets, and so on).

Does this make sense?  I tried the option to import from a Twitter search, but it seems to miss some connections out and doesn't work in the way I propose above (yet still useful).

Maybe it could be another option if this doesn't fit with the current method?

Kind regards, hudmb.


Nov 17, 2009 at 4:44 AM

Thanks for the detailed example.  In the case you've outlined, have all the people in red tweeted the term "NodeXL"?  Or have some of them been mentioned by those who have tweeted "NodeXL" but haven't actually tweeted "NodeXL" themselves?

-- Tony

Nov 17, 2009 at 7:34 AM

You're welcome!

The answer would be the latter.  Those users just *happen* to be mentioned in tweets that contain the target search term.  So, you basically have everything you need in one tweet, as long as that tweet contains the search term, as in the examples above.  In the last example above, the user "seekoeur" makes reference to all those other 10 users in a tweet that contains the target keyword "nodexl".  Nothing more to it than that.

Do you think this option could be easily implemented?

Kind regards, hudmb.

Nov 18, 2009 at 5:04 PM

Then I did understand, and my original comments still hold.

In my view, the feature you are suggesting would break the concept of what the Twitter search network is.  Today, I can look at a graph created by the Import From Twitter Search Network feature and note that every vertex represents a person who has tweeted some term.  That's how the network is defined.  With what you're suggesting, some vertices would represent such people, but others would represent those who were mentioned by people who tweeted the term.  How would a viewer distinguish between the two types of vertices?  I suppose we could add a column called "Vertex Type" and set it to either "Tweeted," "Mentioned by Tweeted," or "Tweeted and Mentioned by Tweeted."  Then, do "Follows" and "Replies-to" edges get added among the "mentioned" vertices, or only among the tweeters?  By this point the whole concept is just too muddled and the design is broken.

So there may be another way for us to provide the information you are seeking, but I don't think it fits into the existing Twitter Search Network.

-- Tony

Nov 18, 2009 at 5:15 PM

OK, cool.

Couple of questions then - what is the current behaviour of the Twitter search network?  You look for all user who Tweeted a certain term but how is the relationship drawn?  The reason I'm asking is that when I did this, there seemed to be many more "mention" links on a regular Twitter search than was extracted using the NodeXL Twitter search, unless I'm misunderstanding how it determines the edges?

Maybe an additional behaviour could be implemented for Twitter "user associations for a given search term, as I propose above?  My objective is to find out the relationship between the person tweeting and whom they @mention in their tweet for a given search term. 

Many thanks, hudmb.

Nov 19, 2009 at 12:23 AM

The Twitter Search Network feature was failing to recognize a reply-to as a mentions, and it was also mistakenly skipping over screen names that ended with a colon, as in "@hudmb:".  That's why Twitter's search was including more mentions than NodeXL's search.  This will be fixed in version Please see

Thanks for pointing out the problem.

-- Tony