Detection of bursty and emerging trends towards identification of researchers at the early stage of trends
Sheron Levar Decker · 2007
Abstract. Detection of trends is important in a variety of areas. Scientific re-search is no exception. While several methods have been proposed for trend de-tection, we argue that there is value on using semantics-based techniques. In particular, we demonstrate the value of using a taxonomy of topics together with data extraction to create a dataset relating publications to topics in the tax-onomy. Compared to other approaches, our method does not have to process the content of the publications. Instead, it uses metadata elements such as keywords and abstracts. Using such dataset, we show that a semantics-based approach can detect “bursty ” and “emerging ” research topic trends. Additionally, our method identifies researchers involved at the early stage of trends. We use known lists of recognized and prolific authors to validate that many of the researchers iden-tified at the early stage of trends have indeed been recognized for their contri-butions on important research trends.