Abstract
Paraphrase detection is important for a number of applications, including plagiarism detection, authorship attribution, question answering, text summarization, text mining in general, etc. In this paper, we give a performance overview of various types of corpus-based models, especially deep learning (DL) models, with the task of paraphrase detection. We report the results of eight models (LSI, TF-IDF, Word2Vec, Doc2Vec, GloVe, FastText, ELMO, and USE) evaluated on three different public available corpora: Microsoft Research Paraphrase Corpus, Clough and Stevenson and Webis Crowd Paraphrase Corpus 2011. Through a great number of experiments, we decided on the most appropriate approaches for text pre-processing: hyper-parameters, sub-model selection—where they exist (e.g., Skipgram vs. CBOW), distance measures, and semantic similarity/paraphrase detection threshold. Our findings and those of other researchers who have used deep learning models show that DL models are very competitive with traditional state-of-the-art approaches and have potential that should be further developed.
Highlights
Paraphrasing is the process of rewriting text to change the form and expression while retaining its original meaning
In this research we compare eight models; namely, LSI, Term-frequency-inverse document frequency (TF-IDF), Word2Vec, Doc2Vec, GloVe, FastText, embeddings from language models (ELMO) [8], and Universal sentence encoder (USE) [9]. We evaluated these models in terms of accuracy, precision, recall, and F1 measure on three different public available corpora: Microsoft Research Paraphrase Corpus (MSRP), Clough and Stevenson (C&S), and Webis Crowd Paraphrase Corpus 2011 (Webis-CPC-11)
Threshold values were calculated from the training parts of datasets and evaluated on the testing parts using the 3-cross validation method for Webis; for MSRP we used predefined train and test datasets; and for C&S we used the 5-cross validation method because of the small corpus and the fact that there are five topics in it
Summary
Paraphrasing is the process of rewriting text to change the form and expression while retaining its original meaning. Automatic paraphrase detection has an important role in the various tasks, including plagiarism detection, authorship attribution, question answering, text summarization, text mining in general, etc. The somewhat more general task of the measuring of the semantic similarity of texts is significant in the domain of natural language processing (NLP). Some of the existing paraphrase systems have performed quite well; there are certain challenges with paraphrase detection. Existing paraphrase systems deliver relatively good results for clean texts, but they do not perform well when applied to noisy texts [1,2,3]. In recent years there was an expansion of deep neural network models’ application to the NLP domain, and that opens up a complete new field for experimentation and improvement of the existing approaches
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
