6 |
Analyzing Non-Textual Content Elements to Detect Academic Plagiarism
|
|
|
|
BASE
|
|
Show details
|
|
7 |
Cross language plagiarism detection with contextualized word embeddings ; Detecção de plágio multilíngue usando word embeddings contextualizadas
|
|
|
|
Abstract:
Plagiarism is the use of someone else’s work without the proper acknowledgment and citation, being one of the most significant publishing issues in academia and science. A study conducted by CopyLeaks in 2020 showed that plagiarism increased by 10% after the transition to online classes during the COVID-19 pandemic. In some cases, authors may translate texts from another language and include them in their work. This more “sophisticated” behavior is known as cross-language plagiarism. In this work, we investigate methods that are used for cross-language plagiarism detection. Although some of the approaches developed until now use word embeddings as part of their pipelines, few explore contextualized word embeddings. Contextualized embeddings can help address fundamental characteristics of language such as polysemy and synonymy by taking into account the context in which a particular word occurs. Pre-trained multilingual models have shown outstanding performance in downstream natural language understanding tasks, such as sentence similarity and next sentence prediction. Motivated by these promising results in tasks related to plagiarism detection, we present a new proposal for cross-language plagiarism detection using pre-trained multilingual models with contextualized embeddings. Experiments performed on different datasets, such as PAN-PC-12, show that the proposed cross-language plagiarism detection using contextualized embeddings outperforms state-of-the-art models by 9% and 11% regarding plagdet results obtained for the English-Spanish and English-German language pairs. ; Plágio é o uso do trabalho de outra pessoa sem o devido reconhecimento e citação, sendo um dos maiores problemas editoriais da academia e da ciência. Um estudo realizado em 2020 pela CopyLeaks mostrou que o plágio aumentou em 10% após a transição para aulas online durante a pandemia da COVID-19. Em alguns casos, os autores podem traduzir textos de outro idioma e incluir em seus próprios trabalhos. Este comportamento mais “sofisticado” é conhecido como plágio multilíngue. Neste trabalho, investigamos métodos que são usados para a detecção do plágio multilíngue. Embora algumas das abordagens desenvolvidas até agora utilizem word embeddings como parte de seu pipeline, poucas delas exploram contexualized word embeddings. Contexualized word embeddings consideram características fundamentais da linguagem, como a polissemia, levando em conta o contexto no qual uma palavra em particular ocorre. Modelos multilíngues pré-treinados têm demonstrado grande desempenho em tarefas multilíngues, tais como similaridade de sentenças e predição de próxima sentença. Assim, com resultados promissores para tarefas relacionadas à detecção de plágio, apresentamos uma nova proposta para a detecção de plágio multilíngue utilizando modelos multilíngues pré-treinados com embeddings contextuais. Experimentos realizados em diferentes conjuntos de dados, como o PAN-PC-12, mostram que a detecção de plágio multilíngue utilizando modelos multilíngues pré-treinados com embeddings contextuais supera supera em 9% e 11% os modelos de última geração em relação aos resultados de plagdet obtidos para os pares de idiomas inglês-espanhol e inglês-alemão.
|
|
Keyword:
BERT; Cross language information retrieval; Cross language plagiarism detection; Plágio; Recuperação de informação : multilíngue; Word embeddings
|
|
URL: http://hdl.handle.net/10183/226141
|
|
BASE
|
|
Hide details
|
|
9 |
Adapting Automatic Summarization to New Sources of Information
|
|
|
|
BASE
|
|
Show details
|
|
10 |
Multilingual Information Access (MLIA) Tools on Google and WorldCat: Bi/Multilingual University Students’ Experience and Perceptions
|
|
|
|
In: FIMS Publications (2019)
|
|
BASE
|
|
Show details
|
|
11 |
Word embeddings for monolingual and cross-language domain-specific information retrieval ; Ordinbäddningar för enspråkig och tvärspråklig domänspecifik informationssökning
|
|
Wigder, Chaya. - : KTH, Skolan för elektroteknik och datavetenskap (EECS), 2018
|
|
BASE
|
|
Show details
|
|
13 |
A Cross-domain and Cross-language Knowledge-based Representation of Text and its Meaning
|
|
|
|
BASE
|
|
Show details
|
|
14 |
On the Feasibility of Character n-Grams Pseudo-Translation for Cross-Language Information Retrieval Tasks
|
|
|
|
BASE
|
|
Show details
|
|
15 |
Studying the Effect and Treatment of Misspelled Queries in Cross-Language Information Retrieval
|
|
|
|
BASE
|
|
Show details
|
|
17 |
A comparative study of online translation services for cross language Information retrieval
|
|
|
|
In: Hosseinzadeh Vahid, Ali, Arora, Piyush orcid:0000-0002-4261-2860 , Liu, Qun orcid:0000-0002-7000-1792 and Jones, Gareth J.F. orcid:0000-0003-2923-8365 (2015) A comparative study of online translation services for cross language Information retrieval. In: 24th International Conference on World Wide Web Companion, 18–22 May 2015, Florence, Italy. ISBN 978-1-4503-3473-0 (2015)
|
|
BASE
|
|
Show details
|
|
18 |
Mining Documents and Sentiments in Cross-lingual Context ; Fouille de documents et d’opinions multilingue
|
|
|
|
In: https://hal.inria.fr/tel-01751251 ; Document and Text Processing. Université de Lorraine, 2015. English. ⟨NNT : 2015LORR0003⟩ (2015)
|
|
BASE
|
|
Show details
|
|
19 |
International Journal Of Web & Semantic Technology (Ijwest) ...
|
|
|
|
BASE
|
|
Show details
|
|
20 |
International Journal Of Web & Semantic Technology (Ijwest) ...
|
|
|
|
BASE
|
|
Show details
|
|
|
|