Establishing a ‘new normal’: detecting fluctuating trends in word frequency over time

Matt Gee, Andrew Kehoe, Antoinette Renouf

    Research output: Chapter in Book/Report/Conference proceedingChapterpeer-review

    Abstract

    In this chapter we introduce statistical methods and associated visualisations for the analysis of lexical change on a monthly basis in a 1.8-billion word news corpus spanning over 30 years. In previous work (Kehoe et al. 2022) we found examples of word frequency change in a data-driven manner by applying existing statistical tests. An ongoing limitation is that, as our diachronic corpus grows, so too does the possibility of a word exhibiting multiple frequency changes in different directions. This chapter reframes the problem as one of time-series segmentation, dividing the frequency history of a word into timespans exhibiting consistent upward or downward change. We then determine reasons for such changes by applying horizon graph visualisations to collocates.
    Original languageEnglish
    Title of host publicationCrossing Boundaries through Corpora. Innovative corpus approaches within and beyond linguistics
    EditorsSarah Buschfeld, Patricia Ronan , Theresa Neumaier, Andreas Weilinghoff, Lisa Westermayer
    PublisherJohn Benjamins Publishing Company
    Chapter6
    Pages125-152
    ISBN (Electronic)9789027246486
    ISBN (Print)9789027215949
    DOIs
    Publication statusPublished (VoR) - 17 Oct 2024

    Publication series

    NameStudies in Corpus Linguistics
    PublisherJohn Benjamins
    Volume119

    Funding

    We thank the Engineering and Physical Science Research Council (EPSRC) for research grants: GR/L08243/01 (APRIL), GR/R16884/01 (WebCorp) and EP/E001300/1 (WebCorpLSE), which funded the infrastructure for this study.

    FundersFunder number
    Engineering and Physical Sciences Research CouncilGR/L08243/01, GR/R16884/01, EP/E001300/1

      Keywords

      • lexical change
      • collocation
      • time series
      • visualisation
      • diachronic change

      Fingerprint

      Dive into the research topics of 'Establishing a ‘new normal’: detecting fluctuating trends in word frequency over time'. Together they form a unique fingerprint.

      Cite this