hub at SemEval-2021 Task 1: Fusion of Sentence and Word Frequency to Predict Lexical Complexity

Proceedings of the 15th International Workshop on Semantic Evaluation (SemEval-2021), pages 598–602 Bangkok, Thailand (online), August 5–6, 2021. ©2021 Association for Computational Linguistics 598 hub at SemEval-2021 Task 1: Fusion of Sentence and Word Frequency to Predict Lexical Complexity Bo Huang, Yang Bai, Xiaobing Zhou* School of Information Science and Engineering Yunnan University, Yunnan, P.R. China *Corresponding author:zhouxb@ynu.edu.com Abstract In this paper, we propose a method of fus- ing sentence information and word frequency information for the SemEval 2021 Task 1- Lexical Complexity Prediction (LCP) shared task. In our system, the sentence informa- tion comes from the RoBERTa model, and the word frequency information comes from the Tf-Idf algorithm. Use Inception block as a shared layer to learn sentence and word fre- quency information. We described the imple- mentation of our best system and discussed our methods and experiments in the task. The shared task is divided into two subtasks. The goal of the two subtasks is to predict the com- plexity of a predetermined word. The evalua- tion index of the task is the Pearson correlation coefficient. Our best performance system has Pearson correlation coefficients of 0.7434 and 0.8000 in the single-token subtask test set and the multi-token subtask test set, respectively. 1 Introduction and Background Language and writing are the main ways we trans- mit knowledge and information. An accurate and efficient understanding of the meaning expressed in the text is of great significance to our learning and production. Vocabulary complexity and reading comprehension are inextricably linked, and overly complex terms may bring bad results (DuBay, 2004). The research of Leroy et al. showed that the use of vocabulary simplification technology is one of the ways to effectively improve readers’ reading comprehension ability (Leroy et al., 2013). Accu- rately predicting lexical complexity can make the system better guide users to use appropriate text, or customize text according to their needs. Especially when some ordinary readers are reading technical text content (Wei et al., 2009). Lexical complexity detection and complex lexical simplification have attracted the attention of the NLP community, and systems have been developed to simplify the text of second language learners (Shardlow, 2014), native speakers with low literacy levels (Specia, 2010), and people with dyslexia (Rello et al., 2013). The topic of the shared task of SemEval 2021 Task 1 is “Lexical Complexity Prediction (LCP)”. The task data set uses English text data in a sin- gle language (Shardlow et al., 2020). There are two subtasks in the task, which are the subtasks for predicting the complexity of a single token and mul- tiple tokens (Shardlow et al., 2021). In this article, we give the method and task result of predicting word complexity. Our system uses a method that combines sentence, word frequency, and context information. The acquisition of sentences and con- text information uses the RoBERTa model (Liu et al., 2019b). The word frequency information comes from the Tf-Idf algorithm (Ramos et al., 2003). The complexity is a continuous value, so the whole task can be …

Content truncated to 500 words