Title: Evolution of prevalence and dominance of HiTEc topics Authors:  Louisa Kontoghiorghes - Kings College London (United Kingdom) [presenting]
Ana Colubi - University of Giessen (Germany)
George Kapetanios - Kings College London (United Kingdom)
Abstract: Text analysis is used to track the evolution of themes within the COST Action HiTEc on a scientific conference's Book of Abstracts (BoAs). To represent HiTEc's relevant themes, a set of keywords is automatically extracted from its proposal. A new topic modeling method is used, the time-varying weighted Latent Dirichlet Allocation (tvwLDA), which estimates the term distribution of the topics and topic distribution of the documents at each time index, enabling the tracking of the topic evolution of the BoAs over time. After applying tvwLDA, the extracted keywords are used to estimate topic prevalence and dominance of HiTEc's themes. The prevalence measures the frequency of HiTEc's themes, while the dominance, a new estimator, combines the topic prevalence with the Simpson index to capture the abundance of HiTEc's related themes in the BoAs.