BERT Embeddings for Automatic Readability Assessment

Proceedings of Recent Advances in Natural Language Processing, pages 44–52 Sep 1–3, 2021. https://doi.org/10.26615/978-954-452-072-4_006 44 Abstract The advancement of the web and information technology has contributed to the rapid growth of digital libraries and automatic machine translation tools which easily translate texts from one language into another. These have increased the content accessible in different languages, which results in easily performing translated plagiarism, which are referred to as “cross-language plagiarism”. Recognition of plagiarism among texts in different languages is more challenging than identifying plagiarism within a corpus written in the same language. This paper proposes a new technique for enhancing English-Arabic cross-language plagiarism detection at the sentence level. This technique is based on semantic and syntactic feature extraction using word order, word embedding and word alignment with multilingual encoders. Those features, and their combination with different machine learning (ML) algorithms, are then used in order to aid the task of classifying sentences as either plagiarized or non-plagiarized. The proposed approach has been deployed and assessed using datasets presented at SemEval-2017. Analysis of experimental data demonstrates that utilizing extracted features and their combinations with various ML classifiers achieves promising results. 1 Introduction The advancement of the Internet and information technology have expanded rapidly the availability of digital libraries and automatic machine translation tools, which facilitate translating a text from one language to another language. This has caused the number of cases of translated plagiarism, referred to as “cross-language plagiarism”, to perform substantially. It is a type of plagiarism that occurs when textual content is translated into another language without giving acknowledgment of original sources. This type of plagiarism is more difficult to detect since each language has its own structure. Several plagiarism detection techniques have been proposed to address monolingual plagiarism, that identify plagiarism instances written in the same language. However, there have been few studies that concentrate on researching and developing methods for identifying cross-language (and in particular English-Arabic) plagiarism. These techniques cannot effectively detect more extensively disguised cases of cross-language plagiarism. Eisa et al. (2015) observed that existing techniques have difficulty detecting linguistic modifications like replacing words and phrases by synonyms. When a text is translated from Arabic into English, synonyms are introduced after the translation, thus it is difficult to identify plagiarism. This paper proposes a new approach for enhancing English-Arabic cross-language plagiarism detection at the sentence level. This technique is based on semantic and syntactic feature extraction using word alignment, word order, word embedding and multilingual encoder models. We investigate the effectiveness of using those features and their combination with different machine learning (ML) algorithms for classifying sentences as either plagiarized or non-plagiarized. The rest of this paper is structured as follows: Section 2 presents a review of related work on cross-language plagiarism detection techniques. In Section 3 we illustrate the proposed approach. The experimental results and discussion are provided in English-Arabic Cross-language Plagiarism Detection Naif Alotaibi Mike Joy Department of Computer Science Department of Computer Science University of Warwick University of Warwick naif.alotaibi@warwick.ac.uk m.s.joy@warwick.ac.uk 45 Section …

Content truncated to 500 words