epsilon-compositionality

Code for our paper Measuring Idiomaticity in Text Embedding Models with 𝜀-compositionality
git clone https://esimon.eu/repos/epsilon-compositionality.git
Log | Files | Refs | Submodules | README

commit 19f2f7818d3862b9d9092c1252e67bf5129b8bf6
parent 1569435e4c9674c02987d325b9baed018bf819f0
Author: Sondre Wold <sondrewold1@gmail.com>
Date:   Mon, 21 Jul 2025 16:31:25 +0200

fix bug

Diffstat:
Mexperiment.py | 4+++-
1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/experiment.py b/experiment.py @@ -353,7 +353,7 @@ class Experiment: def main() -> None: args = parse_args() sep = "\t" - if args.experiment_type in ["ncip", "ncipv2"]: + if args.experiment_type in ["ncip", "ncip-non"]: sep = "," df: pandas.DataFrame = pandas.read_csv(args.data_path, sep=sep) @@ -375,6 +375,8 @@ def main() -> None: C: list[str] = df["both synonym alt3_1"].values.tolist() D: list[str] = df["both synonym alt4_1"].values.tolist() + assert (len(A) + len(B) + len(C) + len(D)) / 4 == len(A) + sentence_A = A + B + C + D sentence_B = B + C + D + A elif args.experiment_type == "sick":