Abstract
Social media communities (e.g. Wikipedia, Flickr, Live Q&A) give rise to distinct types of content, foremost among which are relational content (discussion, chat) and factual content (answering questions, problem-solving). Both users and researchers are increasingly interested in developing strategies that can rapidly distinguish these types of content. While many text-based and structural strategies are possible, we extend two bodies of research that show how social context, and the social roles of answerers can predict content type. We test our framework on a dataset of manually labeled contributions to Microsoft’s Live Q&A and find that it reliably extracts factual and relational messages from the data.
Answer people and discussion people are critical to many social media spaces. Many hosts of communities want to attract and retain them. When these types of participants contribute to a social media space the results can be very positive: questions asked are answered, topics introduced are discussed. But not all participants make a positive contribution, spammers and flamers are often a source of problems in social media spaces. These kinds of contributors also leave distinctive patterns behind:
Author Lines Visualizations of spammers and flame warriors
These patterns illustrate the ways that spammers create only initial turns, never replies. The spammer in the upper left shows a pattern of never skipping a week, and continuously posting just a single message in each thread. In contrast, the remaining Author Lines represent the pattern of heavy discussion people who may verge on flame warriors. The big bubbles indicate the author contributed hundreds of messages to that thread in just one week. The presence of multiple big bubbles indicates the pattern this author has of contributing hundreds of messages to multiple threads over multiple weeks. This high level of focused contribution is an indication of a possible conflict (or perhaps just a spirited debate) and is far different from the pattern of contribution created by answer people or even more moderate discussion people.
Knowing the makeup of the population of your social media space is a key step towards managing these institutions and improving their cllective ability to produce valuable goods and services. What balance of social roles are present in your social media space?
Participants in social media are not all the same. People who contribute to social media come in “flavors” that are created by each participant’s patterns of activity and connection. In the post “Distinguishing social network attributes of online social roles ” some of the social network features of different participants are illustrated. Participants vary in terms of the number of people they connect to and the number of people those people connect to. Some people are connected to many people who are themselves not connected to one another. Other people are connected to many people who do connect to one another. The difference is an indicator of the type of contributions each participant makes.
Another dimension of a social media participant’s activity is their pattern of content creation over time: do people post many messages? Are the messages mostly initial turns, messages that start new threads? Or is the content an author creates mostly replies to other user’s threads? Does the participant contribute many messages in each thread or only a few (or just one)? These different patterns of engagement illustrate the different roles people perform when they interact in a social media space. If you are trying to cultivate an effective online community, knowing about the balance of social roles in your community makes a difference.
Author Lines visualization of Answer People and Discussion People
In this visualization (created by Fernanda Viegas during an internship with me at Microsoft Research several years ago) each image represents a year in the life of a single author to a threaded discussion community (in this case a Usenet newsgroup). Each week in the year is represented as a vertical strip divided in half. In each week a bubble is presented for every thread to which the author contributed. Messages in threads the author started are red and they are placed above the line, messages contributed to threads that were started by any other author appear below the line and are blue. The number of messages contributed to each thread in the given week is represented by the size of the bubble. Bubbles grow more transparent as they grow larger.
In this image, two examples of two different types of authors are displayed. The two authors at the top of this image are “Answer People” who display a pattern of regular contribution to the community discussion throughout the year, although one author (upper right) is increasing and the other decreasing in activity. Both avoid thread initiation (if you look closely you can see a single initial post for the upper left hand author at the end of the year) and both contribute only a few messages to each thread, in most cases only a single message. These brief reply oriented people are mostly answering questions raised by other users. These answer people are the engine of most support oriented discussion spaces. Without them few questions will be answered.
The bottom two authors are discussion people, they both initiate and reply and both contribute several messages to the threads in which they participate. The author in the lower right-hand image has contributed a large volume of messages to several threads over multiple weeks which could indicate a flame war or a vigorous discussion.
