Semantic Similarity Analysis for Paraphrase Identification in Arabic Texts

31st Pacific Asia Conference on Language, Information and Computation (PACLIC 31), pages 274–281 Cebu City, Philippines, November 16-18, 2017 Copyright c⃝2017 Adnen Mahmoud and Mounir Zrigui Semantic Similarity Analysis for Paraphrase Identification in Arabic Texts Adnen Mahmoud LATICE Laboratory Research Department of Computer Science University of Monastir, Tunisia Mahmoud.adnen@gmail.com Mounir Zrigui LATICE Laboratory Research Department of Computer Science University of Monastir, Tunisia Mounir.zrigui@fsm.rnu.tn Abstract Arabic plagiarism detection is a difficult task because of the great richness of Arabic language characteristics of which it is a productive, derivational and inflectional language, on the one hand, and a word can has more than one lexical category in different contexts allows us to have different meanings of the word what changes the meaning of the sentence, on the other hand. In this context, Arabic paraphrase identification allows quantifying how much a suspect Arabic text and source Arabic text are similar based on their contexts. In this paper, we proposed a semantic similarity approach for paraphrase identification in Arabic texts by combining different techniques of Natural Language Processing NLP, such as: Term Frequency- Inverse Document Frequency TF-IDF technique to improve the identification of words that are highly descriptive in each sentence; and distributed word vector representations using word2vec algorithm to reduce computational complexity and to optimize the probability of predicting words in the context given the current center word, which they would be subsequently used to generate a sentence vector representations and after applying a similarity measurement operation based on different metrics of comparison, such as: Cosine Similarity and Euclidean Distance. Finally, our proposed approach was evaluated on the Open Source Arabic Corpus OSAC and obtained a promising rate. 1 Introduction Plagiarism is defined as the unauthorized use or closer imitation for the language and thought of another author and the representation of them that one's own original work based on a set of rules, such as for example: inadequate referencing, direct copy from one or more sources of the text by displacement of words in a sentence, paraphrase and rewrite texts by presenting other's ideas with different words, and translation by expressing an idea in one language into another one (Abderahman et al, 2016) (Gharavi et al, 2016). However, the field of paraphrase detection in Arabic texts is a difficult task because of the great variability of morphological and typographical features of Arabic language where a plagiarized text can include more changes: in the vocabulary, or syntactic, and semantic representation of the text compared with other languages such as Latin or English and we also find that a word can have more than one lexical category in different contexts what changes the meaning of the sentence. Nowadays, Arabic paraphrase identification based on semantic similarity analysis between source text and suspicious text is a difficult task in Natural Language Processing NLP of which we examine the similarity degree of a given pair of texts, in varying in different levels such as word, sentence or paragraph (Vo et al., 2015). Thus, many distributional semantic approaches …

Content truncated to 500 words