CrossLang: the system of cross-lingual plagiarism detection

Plagiarism and text reuse become more available with the Internet development. Therefore it is important to check scientific papers for the fact of cheating, especially in Academia. Existing systems of plagiarism detection show the good performance and have a huge source databases. Thus now it is not enough just to copy the text “as is” from the source document to get the “original” work. Therefore, another type of plagiarism become popular — cross-lingual plagiarism. We present a CrossLang system for such kind of plagiarism detection for English-Russian language pair. The key idea for CrossLang system is that we use the monolingual approach. We have a suspicious Russian document and English reference collection. We reduce the task to one language — we translate the suspicious document into English with the help of machine translation system. After this step we perform the subsequent document analysis. There are two main stages at this analysis: source retrieval stage and document comparison stage. Both of these stages are adapted for our task. At source retrieval stage we need to find the most relevant documents from collection for a given translated suspicious document. Therefore the algorithm is based on aggregation of semantically close words into word classes and thus handles the cases of reformulated passages. The following document comparison is based on phrase embeddings that are trained in unsupervised and semi-supervised regimes. We evaluate CrossLang on the existing and generated datasets. We demonstrate the performance of the whole approach. We integrate the CrossLang in Antiplagiat system (most popular and well-known plagiarism detection system in Russia and CIS) and provide technical characteristics. We also provide the analysis of the system performance. Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. Copyrights for components of this work owned by others than the author(s) must be honored. Abstracting with credit is permitted. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. Request permissions from permissions@acm.org. KDD 2019, August 2019, Anchorage, Alaska USA © 2019 Copyright held by the owner/author(s). Publication rights licensed to ACM. CCS CONCEPTS • Information systems → Information retrieval; • Computing methodologies → Natural language processing; Learning settings; • Applied computing→ Education.