keylexbundles 0.1.3

Compute key lexical bundles (4-grams) keyness between a target corpus and a reference corpus.

keylexbundles

keylexbundles provides a Python function compute_keyness to compute 4-gram lexical bundle keyness between a target and a reference corpus.

---

Install and Usage

First, to install keylexbundles, use pip in terminal:

pip install keylexbundles

Once installed, import the package in your python code by running the following code:

import keylexbundles

Once imported, call the function with two required arguments: target_path and reference_path and an optional arguement output_file:

keylexbundles.compute_keyness(target_path="data/target/", reference_path="data/reference/", output_file="output.csv")

argument	Description
target_path	Folder/directory with .txt files for a target corpus
reference_path	Folder/directory with .txt files for a target corpus
output_file	(Optional) Name of the CSV file to save results to. Defaults to "output.csv".

---

Features

• Accepts two folders that contain .txt files for a target corpus and a reference corpus
• Extracts 4-gram bundles (contiguous sequences of 4 tokens)

Computes:

• Whole-corpus frequency keyness: log-likelihood (G²) using raw whole-corpus token counts
• Text dispersion keyness: log-likelihood (G²) using text dispersion (i.e., the number of texts in which a bundle occurs)
• Mean text frequency keyness: Cohen's d (standardized difference of mean normalized per-text frequencies)

Outputs a CSV with metrics below, sorted primary by text dispersion keyness and secondary by raw whole-corpus token counts in a target corpus

---

Output CSV Columns

Column	Description
lexical bundle	Lexical bundle
whole-corpus frequency keyness	Log-likelihood (G²) based on whole-corpus frequency
text dispersion keyness	Log-likelihood (G²) based on text dispersion
mean text frequency keyness	Cohen's d
raw frequency (target)	token count in target corpus
normed frequency (target)	token frequency per 1,000 words in target corpus
text dispersion (target)	Number of texts where bundle appears in target corpus
mean of normed frequency (target)	Mean per-text normalized frequency in target corpus
sd of normed frequency (target)	Standard deviation of per-text normalized frequency in target corpus
raw frequency (reference)	raw token count in reference corpus
normed frequency (reference)	token frequency per 1,000 words in reference corpus
text dispersion (reference)	Number of texts where bundle appears in reference corpus
mean of normed frequency (reference)	Mean per-text normalized frequency in reference corpus
sd of normed frequency (reference)	Standard deviation of per-text normalized frequency in reference corpus

---

Citation

If you use keylexbundles in your research, please cite it as:
Larsson, T., Kim, T., & Egbert, J. (2025). Introducing and comparing two techniques for key lexical bundles analysis. Research Methods in Applied Linguistics, 4(3), 100245. https://doi.org/10.1016/j.rmal.2025.100245