Lexical Complexity Controlled Sentence Generation

TL;DR

This paper introduces a new task for controlling lexical complexity in sentence generation, proposing a complexity embedding method that improves control and quality in generated sentences, with applications in education and language learning.

Contribution

It presents a novel complexity embedding approach for lexical complexity controlled sentence generation, validated on new datasets in English and Chinese.

Findings

Better lexical complexity control than baselines

Higher quality sentence generation

Effective in both training from scratch and fine-tuning

Abstract

Text generation rarely considers the control of lexical complexity, which limits its more comprehensive practical application. We introduce a novel task of lexical complexity controlled sentence generation, which aims at keywords to sentence generation with desired complexity levels. It has enormous potential in domains such as grade reading, language teaching and acquisition. The challenge of this task is to generate fluent sentences only using the words of given complexity levels. We propose a simple but effective approach for this task based on complexity embedding. Compared with potential solutions, our approach fuses the representations of the word complexity levels into the model to get better control of lexical complexity. And we demonstrate the feasibility of the approach for both training models from scratch and fine-tuning the pre-trained models. To facilitate the research, we develop two datasets in English and Chinese respectively, on which extensive experiments are conducted. Results show that our approach better controls lexical complexity and generates higher quality sentences than baseline methods.