Pars-ABSA: a Manually Annotated Aspect-based Sentiment Analysis Benchmark on Farsi Product Reviews

Proceedings of the 13th Conference on Language Resources and Evaluation (LREC 2022), pages 7056–7060 Marseille, 20-25 June 2022 © European Language Resources Association (ELRA), licensed under CC-BY-NC-4.0 7056 Pars-ABSA: a Manually Annotated Aspect-based Sentiment Analysis Benchmark on Farsi Product Reviews Taha Shangipour Ataei1, Kamyar Darvishi1, Soroush Javdan2, Behrouz Minaei-Bidgoli1, Sauleh Eetemadi1 1Department of Computer Engineering, Iran University of Science and Technology, Tehran, Iran 2School of Computer Science, Carleton University, Ottawa, Canada {taha.pourataei, kamyar.darvishi}@gmail.com soroushjavdan@cmail.carleton.ca {b minaei, sauleh}@iust.ac.ir Abstract Due to the increased availability of online reviews, sentiment analysis witnessed a thriving interest from researchers. Sentiment analysis is a computational treatment of sentiment used to extract and understand the opinions of authors. While many systems were built to predict the sentiment of a document or a sentence, many others provide the necessary detail on various aspects of the entity (i.e., aspect-based sentiment analysis). Most of the available data resources were tailored to English and the other popular European languages. Although Farsi is a language with more than 110 million speakers, to the best of our knowledge, there is a lack of proper public datasets on aspect-based sentiment analysis for Farsi. This paper provides a manually annotated Farsi dataset, Pars-ABSA, annotated and verified by three native Farsi speakers. The dataset consists of 5,114 POSITIVE, 3,061 NEGATIVE and 1,827 NEUTRAL data samples from 5,602 unique reviews. Moreover, as a baseline, this paper reports the performance of some aspect-based sentiment analysis methods focusing on transfer learning on Pars-ABSA. Keywords: Sentiment Analysis, Aspect-based Sentiment Analysis, Opinion Mining, Farsi Language, Persian Language 1. Introduction and Background Nowadays, by being in the era of data explosion where around 500 million tweets are sent daily, and since people are always curious about others’ opinions, one challenge is to build a system to detect and summarize the overall sentiment of these data. Sentiment analy- sis is the computational study of detecting and extract- ing subjective information and attitudes about entities. The entity can represent individuals, events, products, or topics. The output of it is the opinion polarity. Polar- ity is generally expressed in different forms from two classes of POSITIVE and NEGATIVE or three classes of POSITIVE, NEUTRAL, and NEGATIVE, while at some researches, it is represented as a real number between 1-5 stars or 0-10 grade. Sentiment analysis was ac- knowledged in the early 2000s with(Turney, 2002), and (Pang et al., 2002), both of them doing binary clas- sification on reviews. Sentiment analysis is generally performed at three different levels: document-based, sentence-based, and aspect-based. At both the doc- ument and sentence levels of sentiment analysis, the main goal is to detect the polarity of a specific doc- ument or a sentence. In contrast, aspect-based senti- ment analysis is focused on identifying the polarity of the targets expressed in a sentence. A target is an ob- ject, its components, attributes and, features. For in- stance, at (Liu, 2010) a model is provided that identi- fies the polarity of product features that the reviewer has commented on. For example, in …

Content truncated to 500 words