Using Text Similarity to Detect Social Interactions not Captured by Formal Reply Mechanisms

Abstract

In modeling social interaction online, it is important to understand when people are reacting to each other. Many systems have explicit indicators of replies, such as threading in discussion forums or replies and retweets in Twitter. However, it is likely these explicit indicators capture only part of people's reactions to each other; thus, computational social science approaches that use them to infer relationships or influence are likely to miss the mark. This paper explores the problem of detecting non-explicit responses, presenting a new approach that uses tf-idf similarity between a user's own tweets and recent tweets by people they follow. Based on a month's worth of posting data from 449 ego networks in Twitter, this method demonstrates that it is likely that at least 11% of reactions are not captured by the explicit reply and retweet mechanisms. Further, these uncaptured reactions are not evenly distributed between users: some users, who create replies and retweets without using the official interface mechanisms, are much more responsive to followees than they appear. This suggests that detecting non-explicit responses is an important consideration in mitigating biases and building more accurate models when using these markers to study social interaction and information diffusion.

Revealing social networks\' missed behavior: detecting reactions and time-aware analyses

“…This chapter presented a novel method for user's non-explicit reactions to followees' content detection in Twitter. It is based on text similarity scores between a user's tweets and those of their followees [BCJC15]. Our method generates higher scores on average for system tagged Replies and Retweets than Non-Tagged tweets, suggesting that our text similarity scores have the potential to identify users' reactions as stated in our rst research question Ch.2 RQ1.…”

“…1 The contents of this chapter were published at the 2016 ACM 25th International Conference on World Wide Web (WWW'16) [BCJC15]. Some adaptations were made to include further research and adjust the format to this thesis.…”

“…The rst part of this thesis presented a novel method of capturing some of a user's non-explicit reactions to followees' content in Twitter by using text similarity scores between a user's tweets and those of their followees [BCJC15]. The analysis indicates that the method does generate higher scores on average for system tagged Replies and Retweets than Non-Tagged tweets, suggesting that it captures real signal about responses (Ch.2 RQ1).…”

“…The contents of this chapter were published at the 2015 IEEE 11th International Conference on e-Science (e-Science'15)[BCJC15]. Some adaptations were made to include further research and adjust the format to this thesis.2 We evaluate the average score for sets of messages that we know that are reactions.…”

“…The contents of this chapter are as seen in the works published during the course of this Ph.D [BCJC15,BCSC16]…”