The reference corpus matters

Abstract

Abstract This study investigates the effect that reference corpora of different registers have on the content of keyword lists. The study focusses on two target corpora and the keyword lists generated for each when using three distinct reference corpora. The two target corpora consist of published research by faculty at two PhD-granting programs in applied linguistics in North America. The reference corpora comprise published research in applied linguistics, newspaper and magazine articles, and fiction texts, respectively. The findings suggest that while common keywords representing each target corpus emerge regardless of the reference corpus used in the analysis, there are also substantial differences. Primarily, using a reference corpus of the same sub-register as the target corpus better highlights content unique to each target corpus while using a reference corpus of a different register better uncovers words that reflect the register that the target corpora represent. Implications for conducting keyword analysis are discussed.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 18 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.