تحلیل احساس نظرات مردم در مورد خودروهای ملی ایران به کمک مدل زبانی BERT

محورهای موضوعی : پردازش چند رسانه ای، سیستمهای ارتباطی، سیستمهای هوشمند

لیلا گنبدی ¹ , نیلوفر رنجبر ²

1 - دانشجوی کارشناسی ارشد، گروه مهندسی کامپیوتر، دانشگاه خلیج فارس، بوشهر، ایران
2 - مربی، گروه مهندسی کامپیوتر، دانشگاه خلیج فارس، بوشهر، ایران

تاریخ دریافت : 1401/05/19 تاریخ پذیرش : 1402/03/12 تاریخ انتشار : 1401/10/01

کلید واژه: یادگیری ماشین, خودروهای ملی ایران, متن‌کاوی, برت, تحلیل احساس,

چکیده مقاله :

امروزه با پیشرفت شبکه جهانی اینترنت و توسعه شبکه‌های اجتماعی، افراد در مورد موضوعات مختلف سیاسی و اقتصادی به تبادل نظر و گفت‌وگو می‌پردازند. با استفاده از روش‌های متن‌کاوی می‌توان به استخراج اطلاعات از این نظرات پرداخت. یکی از روش‌های تحلیل نظرات، تحلیل نظرات مبتنی بر جنبه است که به کمک آن می‌توان نظرات مردم را در مورد جنبه‌های مختلف یک موضوع تحلیل و بررسی کرد. تحلیل نظرات مردم در مورد صنعت خودروسازی و خودروهای ملی یکی از کاربردهای مهم استفاده از روش‌های متن‌کاوی در موضوعات اقتصادی است. این کار باعث مطلع شدن مدیران صنعت خودروسازی از میزان رضایت واقعی مردم از خودروهای ملی است. برای دست یافتن به این مهم روش‌های مختلفی ارائه شده است. در این مقاله ما به معرفی روشی برای استخراج جنبه‌های مختلف مربوط به خودروهای ملی می‌پردازیم، سپس به کمک مدل زبانی برت برای جملات مختلف، بردار استخراج کرده و پس از آن به کمک شبکه عصبی به دسته‌بندی آن‌ها می‌پردازیم. سپس با داشتن جنبه‌های مختلف مهم برای خودروها و این که نظر مطرح شده به کدام یک از این جنبه‌ها می‌پردازد و حس مطرح شده در نظر مثبت، منفی یا خنثی است، به تحلیل میزان رضایت مردم از جنبه‌های مختلف خودروهای ایرانی می‌پردازیم. هدف نهایی این مقاله تحلیل نظرات مردمی در مورد انواع خودروهای ملی و اجزای آن با هدف بهبود عملکرد خودروسازان است.

چکیده انگلیسی:

Introduction: With the development of the internet and social media, people have been actively discussing political and economic issues, and sharing their opinions online. The vast amount of data generated from these online discussions can be analyzed through text mining methods to extract valuable information. One such method is aspect-based sentiment analysis, which allows for the analysis of people's opinions on various aspects of a topic. In this paper, we will focus on the analysis of people's opinions on the Iranian automobile industry and national cars using aspect-based sentiment analysis.Methodology: To achieve our goal, we first introduce a method for extracting different aspects related to national cars. We then use the "BERT" language model to extract vectors for different sentences related to the various aspects and finally use a neural network to classify the sentiments expressed in these sentences as positive, negative, or neutral.Results: The analysis of public opinions on various aspects of Iranian cars showed that the most discussed aspects were design, quality, and price. The sentiment expressed towards design was largely positive, with people expressing admiration for the unique and modern designs of Iranian cars. The sentiment expressed towards quality was mixed, with some people praising the improved quality of national cars, while others criticized the use of low-quality materials. The sentiment expressed towards price was largely negative, with people complaining about the high prices of Iranian cars compared to their foreign counterparts.Discussion: The results of our analysis provide valuable insights into the level of public satisfaction with various aspects of Iranian cars.  The mixed sentiment expressed towards the quality of Iranian cars highlights the need for manufacturers to focus on using high-quality materials to improve the overall quality of their products. The negative sentiment expressed toward the price of Iranian cars suggests that car manufacturers need to find ways to reduce production costs and offer more competitive pricing. In conclusion, aspect-based sentiment analysis can be used as an effective method to analyze public opinions on various aspects of a topic. Our analysis of public opinions on the Iranian automobile industry and national cars provides valuable insights for car manufacturers to improve the design, quality, and pricing of their products. By taking these insights into account, manufacturers can improve their performance and meet the needs and expectations of their customers.

منابع و مأخذ:

•  Farahani M, Gharachorloo M, Farahani M, Manthouri M. “Parsbert: Transformer-based model for persian language understanding”. Neural Processing Letters. 2021 Dec; 53:3831-47.


•  Blei DM, Ng AY, Jordan MI. “Latent dirichlet allocation”. Journal of machine Learning research. 2003;3(Jan):993-1022.


•  Pai PF, Liu CH. “Predicting vehicle sales by sentiment analysis of twitter data and stock market values”. IEEE Access. 2018 Oct 4; 6:57655-62.


•  Ali F, Kwak D, Khan P, El-Sappagh S, Ali A, Ullah S, Kim KH, Kwak KS. “Transportation sentiment analysis using word embedding and ontology-based topic modeling”. Knowledge-Based Systems. 2019 Jun 15; 174:27-42.


•  Dong J, He F, Guo Y, Zhang H. “A commodity review sentiment analysis based on BERT-CNN model”. In 2020 5th International conference on computer and communication systems (ICCCS), 2020 May 15 (pp. 143-147). IEEE.


•  Wang H, Hu X, Zhang H. “Sentiment analysis of commodity reviews based on ALBERT-LSTM”. In Journal of Physics: Conference Series, 2020 Nov 1 Vol. 1651, No. 1, p. 012022. IOP Publishing.


•  Salari S, Sedighpour N, Vaezinia V, Momtazi S. “Estimation of 2017 Iran’s presidential election using sentiment analysis on social media”. In 2018 4th Iranian Conference on Signal Processing and Intelligent Systems (ICSPIS) 2018 Dec 25 (pp. 77-82). IEEE.


• Heydari M, Khazeni M, Soltanshahi MA. “Deep learning-based sentiment analysis in persian language”. In 2021 7th International Conference on Web Research (ICWR) 2021 May 19 (pp. 287-291). IEEE.


•  Qin Z, Ronchieri E. “Exploring Pandemics Events on Twitter by Using Sentiment Analysis and Topic Modelling”. Applied Sciences. 2022 Nov 22;12(23):11924.


• Jain PK, Quamer W, Saravanan V, Pamula R. “Employing BERT-DCNN with sentic knowledge base for social media sentiment analysis”. Journal of Ambient Intelligence and Humanized Computing. 2022 Jan 25:1-3.


• Maathuis C, Kerkhof I. “The first two months in the war in Ukraine through topic modeling and sentiment analysis.” Regional Science Policy & Practice. 2023 Feb;15(1):56-74.


• Kokab ST, Asghar S, Naz S. “Transformer-based deep learning models for the sentiment analysis of social media data.” Array. 2022 Jul 1;14:100157.


• Mohseni S, Khadivar A, Abassi F. “Sentiment Analysis of Trip Advisor Comments for Iranian Restaurants With a Deep Learning Approach.” Business Intelligence Management Studies,2022, Vol. 10, Issue 40, pp:17-41.


• E, Mohammadi.Sh. “Presenting the model for opinion mining at the document feature level for hotel users' reviews.” Journal of Information and Communication Technology. 2021 Nov;49(49):85.


• Abbasi F, Khadivar A, Yazdinejad M. “Sentiment Analysis of Users for Buying on Cell Phone in Digikala.” BI Management Studies. 2020 Aug 22;8(32):181-210.


• Zare Mehrjardi F, Yazdian-Dehkordi M, Latif A, “Evaluating Machine learning and Deep-learning methods in Sentiment Analysis on Persian Telegram Message”, Soft Computing Journal, 2022 September, Vol.11 , No. 1.


• Tabatabaei S, Barati M, Bayati M. “Sentiment Analysis, Drug Industry, Hybrid Deep Learning, Convolutional And Recursive Architecture.” Intelligent Information Systems Journal. 2022 Dec 10;1(4):21-9.


• Liu X, Li Y, Shao Y, Li A, Liang J. “A sentiment analysis model for car review texts based on adversarial training and whole word mask BERT.” In Intelligent Networked Things: 5th China Conference, CINT 2022, Urumqi, China, August 7-8, 2022, Revised Selected Papers 2023 Jan 28 (pp. 107-121). Singapore: Springer Nature Singapore.


• Sharami JP, Sarabestani PA, Mirroshandel SA. Deepsentipers: Novel deep learning models trained over proposed augmented persian sentiment corpus. arXiv preprint arXiv:2004.05328. 2020 Apr 11.