NCD Classifier is a Python library that implements the method proposed in the paper "Low-Resource" Text Classification: A Parameter-Free Classification Method with Compressors". This method is a non-parametric alternative to deep neural networks for text classification, using a combination of a simple compressor like gzip with a k-nearest-neighbor classifier. It is easy to use, lightweight, and does not require any training parameters, making it suitable for low-resource languages and few-shot settings.
This code was implemented with reference to https://github.com/bazingagin/npc_gzip.
This library is designed with a scikit-learn interface, making it familiar and straightforward for users with experience in scikit-learn.
This library can be installed using pip:
pip install ncd-classifierHere is a simple example of how to use the NCD Classifier:
from ncd_classifier import NCDClassifier
from sklearn.metrics import accuracy_score, confusion_matrix
X_train = ["hello world", "hello?", "world?", "good evening"]
y_train = [0, 0, 0, 1]
X_test = ["hello", "world"]
y_test = [0, 0]
classifier = NCDClassifier(n_jobs=-1, k=3, show_progress=True, label_frequency_weighting=False)
classifier.fit(X_train, y_train)
y_pred = classifier.predict(X_test)
print(y_pred)
print(accuracy_score(y_test, y_pred))
print(confusion_matrix(y_test, y_pred))This code is licensed under the MIT License.
@inproceedings{jiang-etal-2023-low,
title = "{``}Low-Resource{''} Text Classification: A Parameter-Free Classification Method with Compressors",
author = "Jiang, Zhiying and
Yang, Matthew and
Tsirlin, Mikhail and
Tang, Raphael and
Dai, Yiqin and
Lin, Jimmy",
booktitle = "Findings of the Association for Computational Linguistics: ACL 2023",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-acl.426",
pages = "6810--6828",
abstract = "Deep neural networks (DNNs) are often used for text classification due to their high accuracy. However, DNNs can be computationally intensive, requiring millions of parameters and large amounts of labeled data, which can make them expensive to use, to optimize, and to transfer to out-of-distribution (OOD) cases in practice. In this paper, we propose a non-parametric alternative to DNNs that{'}s easy, lightweight, and universal in text classification: a combination of a simple compressor like \textit{gzip} with a $k$-nearest-neighbor classifier. Without any training parameters, our method achieves results that are competitive with non-pretrained deep learning methods on six in-distribution datasets.It even outperforms BERT on all five OOD datasets, including four low-resource languages. Our method also excels in the few-shot setting, where labeled data are too scarce to train DNNs effectively.",
}