Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Research › peer-review
Mapping Opinion Cumulation: Topic Modeling-Based Dynamic Summarization of User Discussions on Social Networks. / Блеканов, Иван Станиславович; Тарасов, Никита Андреевич; Бодрунова, Светлана Сергеевна; Сергеев, Сергей Львович.
Social Computing and Social Media. HCII 2023. Springer Nature, 2023. p. 25–40 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 14025 LNCS).Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Research › peer-review
}
TY - GEN
T1 - Mapping Opinion Cumulation: Topic Modeling-Based Dynamic Summarization of User Discussions on Social Networks
AU - Блеканов, Иван Станиславович
AU - Тарасов, Никита Андреевич
AU - Бодрунова, Светлана Сергеевна
AU - Сергеев, Сергей Львович
N1 - Conference code: 25
PY - 2023
Y1 - 2023
N2 - In the recent years, a lot of methods have been proposed for detection of topicality of user discussions. Recently, the scholars have suggested approaches to tracing topicality evolution, including dynamic topic modeling. However, these approaches are overwhelmingly limited by representation of topics via lists of top words, which only hint to possible contents of topics and does not allow for real mapping of opinion cumulation [1]. We suggest a methodology for discussion mapping that combines neural-network-based encoding of user posts, HDBSCAN-based topic modeling, and abstractive summarization to map large-scale online discussions and trace bifurcation points in opinion cumulation. We test the proposed method on a mid-range dataset on climate change from Reddit and show how discussions may be summarized in a feasible and easily accessible way. Among the rest, we show that the bifurcation points in topicality are often followed by growth of a given topic, which may in future allow for predicting discussion outbursts.
AB - In the recent years, a lot of methods have been proposed for detection of topicality of user discussions. Recently, the scholars have suggested approaches to tracing topicality evolution, including dynamic topic modeling. However, these approaches are overwhelmingly limited by representation of topics via lists of top words, which only hint to possible contents of topics and does not allow for real mapping of opinion cumulation [1]. We suggest a methodology for discussion mapping that combines neural-network-based encoding of user posts, HDBSCAN-based topic modeling, and abstractive summarization to map large-scale online discussions and trace bifurcation points in opinion cumulation. We test the proposed method on a mid-range dataset on climate change from Reddit and show how discussions may be summarized in a feasible and easily accessible way. Among the rest, we show that the bifurcation points in topicality are often followed by growth of a given topic, which may in future allow for predicting discussion outbursts.
KW - Abstractive summarization
KW - Cumulative deliberation
KW - Dynamic summarization
KW - Mapping discussions
KW - Opinion cumulation
KW - Topic modeling
UR - https://link.springer.com/chapter/10.1007/978-3-031-35915-6_3
UR - https://www.mendeley.com/catalogue/25e9215a-a7e8-3e3c-abb0-3308d6ffea17/
U2 - 10.1007/978-3-031-35915-6_3
DO - 10.1007/978-3-031-35915-6_3
M3 - Conference contribution
SN - 978-3-031-35914-9
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 25
EP - 40
BT - Social Computing and Social Media. HCII 2023
PB - Springer Nature
T2 - 25TH INTERNATIONAL CONFERENCE ON HUMAN-COMPUTER INTERACTION
Y2 - 23 July 2023 through 28 July 2023
ER -
ID: 110777051