Learning Syntactic Dense Embedding with Correlation Graph for Automatic Readability Assessment

TL;DR

This paper introduces a method to enhance neural readability assessment models by integrating linguistic features through syntactic dense embeddings learned via a correlation graph, improving performance over BERT-only models.

Contribution

It presents a novel approach to incorporate linguistic features into neural models using correlation graphs to learn syntactic embeddings, boosting readability assessment accuracy.

Findings

Enhanced readability assessment performance with the proposed method

Correlation graph effectively captures feature relationships

Method outperforms BERT-only models on multiple datasets

Abstract

Deep learning models for automatic readability assessment generally discard linguistic features traditionally used in machine learning models for the task. We propose to incorporate linguistic features into neural network models by learning syntactic dense embeddings based on linguistic features. To cope with the relationships between the features, we form a correlation graph among features and use it to learn their embeddings so that similar features will be represented by similar embeddings. Experiments with six data sets of two proficiency levels demonstrate that our proposed methodology can complement BERT-only model to achieve significantly better performances for automatic readability assessment.