International Journal of Data Science and Analytics Информационные системы
A cross-domain topic recommendation system for enhancing media-academic exchange in cybersecurity
Аннотация
The exchange of knowledge and information among different domains is of paramount importance in this digital era. However, there usually exists a language gap between the two parties. For example, the academic domain contains heavily technical content which are not easily absorbed by journalists. In contrast, the media domain includes content that is not related to a particular research topic, making it hard for an academic researcher to find research problems to solve. In this research, we build a cross-domain topic recommendation system (CDTRS) to find the cross-domain linkage between the academic domain and the media domain. We show how we bridge the knowledge and language gap between the academic and media domains by first exploring topics that arise in both domains. Then, we build a cross-domain similar topic network (CDSTN) with nodes representing topics from two domains produced by topic modeling to link similar topics. Next, we employ a neural embedding approach, Node2Vec, to predict extended connections and recommend topics from one domain to another. We use a case of cybersecurity to show a step-by-step illustration of our recommendation system. We use cybersecurity-oriented content in media domains from the New York Times and academic domain from highly reputable cybersecurity academic conferences between 2008 and 2018. We show the effectiveness of our proposed recommendation system for recommending topics from the academic domain to the media domain and vice versa and find that our recommendation system can perform well even with limited training data. Our semantic-based and structure-based cross-domain recommendation system can provide insightful information for academic scholars, journalists, and the general public. It can be adopted in other domains with language gaps.