Elamur Research⌕
☰
home / papers / 2021 / Interactive Learning Approach for Arabic Target-Based Sentiment Analysis

Interactive Learning Approach for Arabic Target-Based Sentiment Analysis

Paper · 2021 · Husamelddin Balla, Marisa Llorens Salvador and Sarah Jane Delany

Summary

The paper tackles target-based sentiment analysis (TBSA) for Arabic. The task is to assign a polarity to each specific aspect or entity mentioned in a review, not to the whole sentence. Arabic TBSA has received little attention compared with English. The authors propose IA-LSTM, which extends the English IAN model of Ma et al. (2017). IAN models the target and only one context. IA-LSTM gives the left context, the target and the right context their own LSTM encoders, and three attention networks let the target and each context attend to one another. The motivation is Arabic's right-to-left script: opinion words that follow a target appear to its left, and IAN ignored that side. IA-LSTM was tested on an Arabic hotel review dataset (from SemEval-2016) and an Arabic book review dataset. It reached 83.10 accuracy on hotels, slightly above the previous best, AB-LSTM-PC (82.60). An ablation confirms that interaction with both sides helps. This is a modest but clear gain for a language with few resources, and the code is released on GitHub.

Key points

  • IA-LSTM is built from three LSTM encoders (left context, target, right context) plus interactive attention between the target and each context. The combined representation is fed to a softmax classifier.
  • Hotel reviews accuracy:
    • IA-LSTM: 83.10
    • INSIGHT-1: 82.7
    • AB-LSTM-PC: 82.60
    • IAN: 81.90
    • Plain LSTM: 74.26
    • SVM n-gram baseline: 76.40
  • Ablation (hotel / book accuracy):
    • No interaction: 76.75 / 74.64
    • Full model: 83.10 / 80.82
    • One-sided variants fall in between, and the paper reports the left-side variant as the stronger one.
  • Per-class accuracy shows the effect of the imbalanced data. Neutral is weakest (76.14 hotel, 65.82 book), while positive and negative score in the mid-80s or higher.
  • INSIGHT-1 and AB-LSTM-PC results are quoted from earlier publications. LSTM and IAN were re-implemented and run by the authors.

Interesting ideas

  • I1. Each sentence is split into three spans around the opinion target, each with its own encoder, and the target and its contexts are allowed to condition each other's attention. This anchors the evaluation on a specific entity rather than the whole sentence.
  • I2. The authors chose which context to add by reasoning about where evaluative words typically sit given the script direction and word order. This is a linguistically motivated design choice rather than blind architecture search.
  • I3. The ablation switches interaction off entirely, then enables it one side at a time, then both. This isolates how much each information channel contributes.
  • I4. Accuracy is reported per polarity class because the classes are imbalanced, which exposes that the rare neutral class is much weaker than the headline number suggests.

For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

  • I1 (frames): attitude or evaluation frames in prag could score evaluative words relative to a chosen target entity, using the target's dependents and the material on each side. That yields per-entity polarity with an explainable link to the words responsible.
  • I2 (tree): when picking feature windows for the perceptron parser or gparse, we could check where modifiers and predicates of a head usually occur in English. We would then weight or extend context on that side instead of using symmetric windows.
  • I3 (learn): the same "disable one channel" scheme can work as a negative control for any learned component. Removing left or right context features, or a witness parser, should measurably hurt if the component truly uses it.
  • I4 (learn): our evaluation reports could always break scores down by label, such as rare UD relations or rare frame roles. That way overall accuracy does not hide weak minority categories.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🎓Technological University DublinDublin · Ireland6 works · 188 citations

 Fields, topics and concepts

📚 Shelves 6The field of science this work belongs to (OpenAlex fields).Computer Science · 18,614 worksArtificial Intelligence · 17,965 worksArts and Humanities · 17,450 worksLanguage and Linguistics · 17,296 worksDecision Sciences · 249 worksManagement Science and Operations Research · 134 works

🧭 Topics 2Research subjects and languages the work is about.Sentiment & emotion · 1,035 worksSemitic languages · 3,066 works

Cite

Cite as Husamelddin Balla, Marisa Llorens Salvador and Sarah Jane Delany. 2021. Interactive Learning Approach for Arabic Target-Based Sentiment Analysis. Proceedings of the International Conference on Recent Advances in Natural Language Processing (RANLP 2021).

The paper itself, not this page. BibTeX from the library's record of the work.

@inproceedings{balla2021interactive,
  title = {{Interactive Learning Approach for Arabic Target-Based Sentiment Analysis}},
  author = {Husamelddin Balla, Marisa Llorens Salvador and Sarah Jane Delany},
  booktitle = {Proceedings of the International Conference on Recent Advances in Natural Language Processing (RANLP 2021)},
  year = {2021},
  doi = {10.26615/978-954-452-072-4\_014},
  url = {https://doi.org/10.26615/978-954-452-072-4_014}
}

References

35 references, 7 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Muhammad Abdul-Mageed and Mona T Diab. 2012. Awatif: A multi-genre corpus for modern standard arabic subjectivity and sentiment analysis. In LREC, volume 515, pages 3907–3914.
  2. Khurshid Ahmad. 2006. Multi-lingual sentiment analysis of financial news streams. PoS, page 001. find
  3. Hamed Al-Rubaiee, Renxi Qiu, and Dayou Li. 2016. Identifying mubasher software products through sentiment analysis of arabic tweets. In 2016 International Conference on Industrial Informatics and Computer Systems (CIICS), pages 1–6. IEEE. find
  4. Mohammad Al-Smadi, Omar Qawasmeh, Mahmoud Al-Ayyoub, Yaser Jararweh, and Brij Gupta. 2018. Deep recurrent neural network vs. support vector machine for aspect-based sentiment analysis of arabic hotels’ reviews. Journal of computational science, 27:386–393. ↗
  5. Mohammad Al-Smadi, Omar Qawasmeh, Bashar Talafha, and Muhannad Quwaider. 2015. Human annotated arabic dataset of book reviews for aspect based sentiment analysis. In 2015 3rd International Conference on Future Internet of Things and Cloud, pages 726–730. IEEE. find
  6. Mohammad Al-Smadi, Bashar Talafha, Mahmoud Al-Ayyoub, and Yaser Jararweh. 2019. Using long short-term memory deep neural networks for aspect-based sentiment analysis of arabic reviews. International Journal of Machine Learning and Cybernetics, 10(8):2163–2175. ↗
  7. Yousif Almas and Khurshid Ahmad. 2007. A note on extracting ‘sentiments’ in financial news in english, arabic & urdu. In The Second Workshop on Computational Approaches to Arabic Script-based Languages, pages 1–12. find
  8. Naaima Boudad, Rdouan Faizi, Rachid Oulad Haj Thami, and Raddouane Chiheb. 2018. Sentiment analysis in arabic: A review of the literature. Ain Shams Engineering Journal, 9(4):2479–2490. ↗
  9. Abdelghani Dahou, Shengwu Xiong, Junwei Zhou, Mohamed Houcine Haddoud, and Pengfei Duan. 2016. Word embeddings and convolutional neural network for arabic sentiment classification. In Proceedings of coling 2016, the 26th international conference on computational linguistics: Technical papers, pages 2418–2427.
  10. Ashraf Elnagar, Ridhwan Al-Debsi, and Omar Einea. 2020. Arabic text classification using deep learning models. Information Processing & Management, 57(1):102121. ↗
  11. Noura Farra, Elie Challita, Rawad Abou Assi, and Hazem Hajj. 2010. Sentence-level and document-level sentiment mining for arabic texts. In 2010 IEEE international conference on data mining workshops, pages 1114–1119. IEEE. find
  12. Felix A Gers, Jürgen A Schmidhuber, and Fred A Cummins. 2000. Learning to forget: Continual prediction with lstm. Neural Computation, 12(10):2451– 2471. ↗
  13. Minqing Hu and Bing Liu. 2004. Mining and summarizing customer reviews. In Proceedings of the tenth ACM SIGKDD international conference on Knowledge discovery and data mining, pages 168–177. find
  14. Long Jiang, Mo Yu, Ming Zhou, Xiaohua Liu, and Tiejun Zhao. 2011. Target-dependent twitter sentiment classification. In Proceedings of the 49th annual meeting of the association for computational linguistics: human language technologies, pages 151–160. find
  15. Xin Li, Lidong Bing, Piji Li, Wai Lam, and Zhimou Yang. 2018. Aspect term extraction with history attention and selective transformation. arXiv preprint arXiv:1805.00760. ↗
  16. Dehong Ma, Sujian Li, Xiaodong Zhang, and Houfeng Wang. 2017. Interactive attention networks for aspect-level sentiment classification. arXiv preprint arXiv:1709.00893. ↗
  17. Yukun Ma, Haiyun Peng, and Erik Cambria. 2018. Targeted aspect-based sentiment analysis via embedding commonsense knowledge into an attentive lstm. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 32. find
  18. AL-Smadi Mohammad, Omar Qwasmeh, Bashar Talafha, Mahmoud Al-Ayyoub, Yaser Jararweh, and Elhadj Benkhelifa. 2016. An enhanced framework for aspect-based sentiment analysis of hotels’ reviews: Arabic reviews case study. In 2016 11th International Conference for Internet Technology and Secured Transactions (ICITST), pages 98–103. IEEE. find
  19. Rehab Duwairi and Mahmoud El-Orfali. 2014. A study of the effects of preprocessing strategies on sentiment analysis for arabic text. Journal of Information Science, 40(4):501–513. ↗
  20. Mohammad Erfan Mowlaei, Mohammad Saniee Abadeh, and Hamidreza Keshavarz. 2020. Aspect-based sentiment analysis using adaptive aspect-based lexicons. Expert Systems with Applications, 148:113234. ↗
  21. Mohamed Elhawary and Mohamed Elfeky. 2010. Mining arabic business reviews. In 2010 ieee international conference on data mining workshops, pages 1108–1113. IEEE. find
  22. Mahmoud Nabil, Mohamed Aly, and Amir Atiya. 2015. Astd: Arabic sentiment tweets dataset. In Proceedings of the 2015 conference on empirical methods in natural language processing, pages 2515–2519. ↗
  23. Maria Pontiki, Dimitrios Galanis, Haris Papageorgiou, Ion Androutsopoulos, Suresh Manandhar, Mohammad Al-Smadi, Mahmoud Al-Ayyoub, Yanyan Zhao, Bing Qin, Orphée De Clercq, et al. 2016. Semeval-2016 task 5: Aspect based sentiment analysis. In International workshop on semantic evaluation, pages 19–30. ↗
  24. Ana-Maria Popescu and Orena Etzioni. 2007. Extracting product features and opinions from reviews. In Natural language processing and text mining, pages 9–28. Springer. ↗
  25. Ning Qian. 1999. On the momentum term in gradient descent learning algorithms. Neural networks, 12(1):145–151. ↗
  26. Guang Qiu, Bing Liu, Jiajun Bu, and Chun Chen. 2009. Expanding domain sentiment lexicon through double propagation. In IJCAI, volume 9, pages 1199– 1204. Citeseer. find
  27. Sebastian Ruder, Parsa Ghaffari, and John G Breslin. 2016. Insight-1 at semeval-2016 task 5: Deep learning for multilingual aspect-based sentiment analysis. arXiv preprint arXiv:1609.02748. ↗
  28. Mike Schuster and Kuldip K Paliwal. 1997. Bidirectional recurrent neural networks. IEEE transactions on Signal Processing, 45(11):2673–2681. ↗
  29. Amira Shoukry and Ahmed Rafea. 2012. Sentence-level arabic sentiment analysis. In 2012 International Conference on Collaboration Technologies and Systems (CTS), pages 546–550. IEEE. find
  30. Abu Bakr Soliman, Kareem Eissa, and Samhaa R El-Beltagy. 2017. Aravec: A set of arabic word embedding models for use in arabic nlp. Procedia Computer Science, 117:256–265. find
  31. Youwei Song, Jiahai Wang, Tao Jiang, Zhiyue Liu, and Yanghui Rao. 2019. Attentional encoder network for targeted sentiment classification. arXiv preprint arXiv:1902.09314. ↗
  32. Aleš Tamchyna and Kateřina Veselovská. 2016. Ufal at semeval-2016 task 5: recurrent neural networks for sentence classification. In Proceedings of the 10th international workshop on semantic evaluation (SEMEVAL-2016), pages 367–371. find
  33. Duyu Tang, Bing Qin, Xiaocheng Feng, and Ting Liu. 2015. Effective lstms for target-dependent sentiment classification. arXiv preprint arXiv:1512.01100. ↗
  34. Wei Xue and Tao Li. 2018. Aspect based sentiment analysis with gated convolutional networks. arXiv preprint arXiv:1805.07043. ↗
  35. Li Zhuang, Feng Jing, and Xiao-Yan Zhu. 2006. Movie review mining and summarization. In Proceedings of the 15th ACM international conference on Information and knowledge management, pages 43–50. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/balla-2021-interactive-learning-approach-arabic-target/