Recommending interesting connections among tweeters (incl. demo)

It is an interesting issue how social media systems can be applied to the supporting of serendipity. As a popular social media system, Twitter allows people to rapidly spread and exchange messages, and we assume it can be good a information source for exploring this issue. Although it was mainly used for personal chatting at the beginning, we have seen an increasing use of it for more serious purposes, such as research and business. Therefore, it now provides rich up-to-date information for finding a variety of interesting connections in research and work in addition to daily life.

One application scenario of twitter information is to promote interaction among participants in public events by recommending people connections, which may potentially lead to serendipity. Events like conferences often involve quite a large group of people, and it is a tricky business to identify people from the crowd who share something interesting with you. If we can suggest such connections on the spot in timely manner, it may trigger interesting interactions among people which may lead to serendipitous encounters. Twitter provides an ideal information source for it due to its real-time nature, although it only applies to Twitter users. As the increase of the Twitter user base, nevertheless, we can expect a wide coverage of such methods in near future.

There is a couple of important requirements to be considered when we design such a system. For example, such a system should not require substantial amount of input from users. Ideally, after an initial starting-up process, which is to be as simple as possible, the system should be mainly running in the background and minimize interruption to the users. In the above scenario, the users only need to include a pre-defined hashtag in their tweets, or simply allow the system to collect all of their tweets and forget about the hashtag. After that, the system collects and analyses the users’ data behind the scene and prepares recommendations which are available and accessible to the users when and where they want it.

A trial system has been developed to explore the feasibility of such an application of Twitter in a public event scenario. The system is implemented as a cluster of components carrying out functionalities including collecting, analysing, connecting and presenting connections to users, together of which provide a web service, as illustrated by the diagram below:

In a nutshell, the system keeps monitoring and collecting the twitter updates of whom are registered to the system. A content analysis component keeps extracting representative terms as summary of interests for each of the users, employing a cluster of NLP (Natural Language Processing) services. Meanwhile, another component keeps detecting users who share some interests. The results can be presented to the users in various ways, from simple tabular lists to complex graphical visualization. For this trial system, we use a graph visualization based on Google Graph API, as illustrated below:

This system was demonstrated in the DE Summer School DocFest and attracted interesting feedbacks from the audience. As a trial system under development, it has limited functionalities for now, but it demonstrates the possibility of supporting serendipity based on Twitter. As an ongoing work, more news to follow, so please keep your eyes on this site.

You are welcome to try the live demo system following this this link (reduce the font size to see the whole graph). All you need to do is to include hashtag “#srnmsg” in your tweets to join the connecting exercise. (It is live unless the server crashes – quite possible for a prototype; let us know if it happens). Have a fun!

Scott Piao
School of Computing and Communications
Lancaster University
Twitter ID: @scottpiao