Arabic Sentiment analysis: an empirical study of machine translation's impact

Amira Barhoumi 1 Chafik Aloulou Nathalie Camelin 1 Yannick Estève 1 Lamia Belguith 2
1 LIUM - Laboratoire d'Informatique de l'Université du Mans
2 MIRACL - Multimedia, InfoRmation systems and Advanced Computing Laboratory
Abstract : The largest amount of Sentiment Analysis has been carried out for English language. To deal with Arabic sentiment analysis, machine translation of English resources or Arabic texts may be applied to built Arabic sentiment analysis systems. In this paper, we translate Arabic dataset into English and study the impact of machine translation while considering a standard Arabic system as a baseline. Experiments show that sentiment analysis of Arabic content translated into English reach a competitive performance with respect to standard sentiment analysis of Arabic texts. This suggests that machine translation can successfully transfer the expression of sentiment or polarity. Moreover , we explored the multi-domain extending of training data in order to enhance performance and we show that we should have, in the training set, data whose domain is the same as the domain of evaluation dataset.
Keywords : Sentiment analysis opinion mining machine translation document embeddings machine learning Arabic language
Document type :
Conference papers
Complete list of metadatas

Cited literature [16 references]  Display  Hide  Download
https://hal.archives-ouvertes.fr/hal-02042313
Contributor : Amira Barhoumi <>
Submitted on : Wednesday, February 20, 2019 - 12:22:30 PM
Last modification on : Sunday, June 2, 2019 - 10:24:02 AM

File

LPKM2018_paper_12.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-02042313, version 1

Collections

LIUM | LIUM-LST | UNIV-LEMANS

Citation

Amira Barhoumi, Chafik Aloulou, Nathalie Camelin, Yannick Estève, Lamia Belguith. Arabic Sentiment analysis: an empirical study of machine translation's impact. LANGUAGE PROCESSING AND KNOWLEDGE MANAGEMENT INTERNATIONAL CONFERENCE (LPKM2018), Oct 2018, Sfax, Tunisia. ⟨hal-02042313⟩

Export

BibTeX TEI DC DCterms EndNote

Share

Metrics

Record views

91

Files downloads

62

Contact

support.ccsd.cnrs.fr
hal.support@ccsd.cnrs.fr
Logo CCSD