To determine how closely a Twitter user kept to a single topic, Yi-Chia used a statistical method called average pairwise cosine similarity, or AvgCosSim, which measures the degree to which multiple sets of text, such as groups of Twitter messages, share common words. When people talk about a single topic, they tend to use the same words, Yi-Chia explained. U.S. conservatives discussing the topic of health care, for instance, frequently use words such as "Obama," "Obamacare," "socialism," and "repeal" for instance. The higher the AvgCosSim score, the more likely the Twitter messages being analyzed were on a single topic.
In her analysis, Yi-Chia found that Twitters users with higher AvgCosSim scores had more followers than those with lower scores. "Topic focus is significantly correlated with membership size," Yi-Chia said. In other words, Twitter users who stuck to a single topic gathered more followers and created a more vibrant community than those who touched upon multiple topics. For instance, increasing the AvgCosSim score by 0.01 (in effect moving from the 25th to the 50th percentile) on average attracted 111 more followers.
The audience in attendance praised the work, noting that it pointed to future areas of research. Someone noted that some Twitter accounts reaped many followers not by staying on one topic, but by some other means, such as offering only humorous Twitter messages. Yi-Chia admitted that the study does not take into account such types of Twitter accounts, though the study did factor out celebrities, such as Ashton Kutcher, who gain large followings through name recognition.