Saved in:
Bibliographic Details
Main Authors: Xu, Qingshu, Jiao, Hong, Zhou, Tianyi, Li, Ming, Zhang, Nan, Peters, Sydney, Fu, Yanbin
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2510.05129
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911203458023424
author Xu, Qingshu
Jiao, Hong
Zhou, Tianyi
Li, Ming
Zhang, Nan
Peters, Sydney
Fu, Yanbin
author_facet Xu, Qingshu
Jiao, Hong
Zhou, Tianyi
Li, Ming
Zhang, Nan
Peters, Sydney
Fu, Yanbin
contents Accurate alignment of items to content standards is critical for valid score interpretation in large-scale assessments. This study evaluates three automated paradigms for aligning items with four domain and nineteen skill labels. First, we extracted embeddings and trained multiple classical supervised machine learning models, and further investigated the impact of dimensionality reduction on model performance. Second, we fine-tuned eight BERT model and its variants for both domain and skill alignment. Third, we explored ensemble learning with majority voting and stacking with multiple meta-models. The DeBERTa-v3-base achieved the highest weighted-average F1 score of 0.950 for domain alignment while the RoBERTa-large yielded the highest F1 score of 0.869 for skill alignment. Ensemble models did not surpass the best-performing language models. Dimension reduction enhanced linear classifiers based on embeddings but did not perform better than language models. This study demonstrated different methods in automated item alignment to content standards.}
format Preprint
id arxiv_https___arxiv_org_abs_2510_05129
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Automated Alignment of Math Items to Content Standards in Large-Scale Assessments Using Language Models
Xu, Qingshu
Jiao, Hong
Zhou, Tianyi
Li, Ming
Zhang, Nan
Peters, Sydney
Fu, Yanbin
Computation and Language
Machine Learning
Accurate alignment of items to content standards is critical for valid score interpretation in large-scale assessments. This study evaluates three automated paradigms for aligning items with four domain and nineteen skill labels. First, we extracted embeddings and trained multiple classical supervised machine learning models, and further investigated the impact of dimensionality reduction on model performance. Second, we fine-tuned eight BERT model and its variants for both domain and skill alignment. Third, we explored ensemble learning with majority voting and stacking with multiple meta-models. The DeBERTa-v3-base achieved the highest weighted-average F1 score of 0.950 for domain alignment while the RoBERTa-large yielded the highest F1 score of 0.869 for skill alignment. Ensemble models did not surpass the best-performing language models. Dimension reduction enhanced linear classifiers based on embeddings but did not perform better than language models. This study demonstrated different methods in automated item alignment to content standards.}
title Automated Alignment of Math Items to Content Standards in Large-Scale Assessments Using Language Models
topic Computation and Language
Machine Learning
url https://arxiv.org/abs/2510.05129