Saved in:
Bibliographic Details
Main Authors: Zhang, Duo, Mo, Junyi
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2506.21848
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913970987728896
author Zhang, Duo
Mo, Junyi
author_facet Zhang, Duo
Mo, Junyi
contents Deep learning has significantly advanced NLP, but its reliance on large black-box models introduces critical interpretability and computational efficiency concerns. This paper proposes LinguaSynth, a novel text classification framework that strategically integrates five complementary linguistic feature types: lexical, syntactic, entity-level, word-level semantics, and document-level semantics within a transparent logistic regression model. Unlike transformer-based architectures, LinguaSynth maintains interpretability and computational efficiency, achieving an accuracy of 84.89 percent on the 20 Newsgroups dataset and surpassing a robust TF-IDF baseline by 3.32 percent. Through rigorous feature interaction analysis, we show that syntactic and entity-level signals provide essential disambiguation and effectively complement distributional semantics. LinguaSynth sets a new benchmark for interpretable, resource-efficient NLP models and challenges the prevailing assumption that deep neural networks are necessary for high-performing text classification.
format Preprint
id arxiv_https___arxiv_org_abs_2506_21848
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle LinguaSynth: Heterogeneous Linguistic Signals for News Classification
Zhang, Duo
Mo, Junyi
Computation and Language
Deep learning has significantly advanced NLP, but its reliance on large black-box models introduces critical interpretability and computational efficiency concerns. This paper proposes LinguaSynth, a novel text classification framework that strategically integrates five complementary linguistic feature types: lexical, syntactic, entity-level, word-level semantics, and document-level semantics within a transparent logistic regression model. Unlike transformer-based architectures, LinguaSynth maintains interpretability and computational efficiency, achieving an accuracy of 84.89 percent on the 20 Newsgroups dataset and surpassing a robust TF-IDF baseline by 3.32 percent. Through rigorous feature interaction analysis, we show that syntactic and entity-level signals provide essential disambiguation and effectively complement distributional semantics. LinguaSynth sets a new benchmark for interpretable, resource-efficient NLP models and challenges the prevailing assumption that deep neural networks are necessary for high-performing text classification.
title LinguaSynth: Heterogeneous Linguistic Signals for News Classification
topic Computation and Language
url https://arxiv.org/abs/2506.21848