Detecting bursty terms in computer science research

Abstract

Research topics rise and fall in popularity over time, some more swiftly than others. The fastest rising topics are typically called bursts; for example "deep learning", "internet of things" and "big data". Being able to automatically detect and track bursty terms in the literature could give insight into how scientific thought evolves over time. In this paper, we take a trend detection algorithm from stock market analysis and apply it to over 30 years of computer science research abstracts, treating the prevalence of each term in the dataset like the price of a stock. Unlike previous work in this domain, we use the free text of abstracts and titles, resulting in a finer-grained analysis. We report a list of bursty terms, and then use historical data to build a classifier to predict whether they will rise or fall in popularity in the future, obtaining accuracy in the region of 80%. The proposed methodology can be applied to any time-ordered collection of text to yield past and present bursty terms and predict their probable fate.

Topic space trajectories

“…A surprising result of this work different from ours, was that clustering and classification were determined as a single topical cluster. Tattershall et al (2020) analyze computer science research through burst detection and found some results similar to ours, e.g., the rise of popularity of Social Networks and related terms between 2004 and 2014, the lowered interest in web 2.0 in recent years. In Bhattacharya (2019), an analysis of machine learning literature has been performed based on measures from social network analysis, VOSviewer and keywords from the web of science.…”

“…A field sometimes using this approach and related to our work is the one of burst detection, i.e., the detection of sudden increases or decreases of interest in a topic (Kleinberg 2002). As a more recent application Tattershall et al (2020) applies burst detection methods from stock market trend analysis to detect bursts in research topics based on normalized term frequencies. A disadvantage over our approach here is the termbased analysis, which requires additional manual work and does not give an overall summary.…”

55 years of Abacus: Evolution of Research Streams and Future Research Directions

“…The algorithm uses information about topics in the previous step of the analysis to identify periods during which there was an increased number of articles published on a certain topic. Prior research (e.g., Tattershall et al ., 2020) has developed various burst detection algorithms. Here, we base our analysis on Kleinberg's (2003) burst detection algorithm, which has been used in prior publications to successfully identify emerging topics in academic publications (e.g., Mane and Börner, 2004).…”

Sixty years of Accounting & Finance: a bibliometric analysis of major research themes and contributions

“…In order to determine the ‘hot’ topics, our research uses burst detection algorithms to identify periods of increased presence of topics. While there are a number of different burst detection algorithms (see Tattershall et al ., 2020), one of the most popular is Kleinberg’s (2003) burst detection algorithm, as it is widely used to identify emerging topics from increases in their occurrence in a literature (e.g., Mane and Börner, 2004). To implement the burst detection algorithm, we used the Python library called burst_detection (Version 0.1.3).…”