International Journal of Hybrid Intelligent Systems -1 (2020) 1–12 1
DOI 10.3233/HIS-200285
IOS Press
A hybrid method for Arabic aspect-based
sentiment analysis
Sana Trigui
a,*
, Ines Boujelben
a,b
, Salma Jamoussi
a,c
and Yassine BenAyed
a,c
a
Miracl, University of Sfax, Sfax, Tunisia
b
Higher Institute of Computer Science and Multimedia of Gabes, Gabes, Tunisia
c
Higher Institute of Computer Science and Multimedia of Sfax, Sfax, Tunisia
Abstract Nowadays, sentiment analysis has been a very active research area with the increase of social media data. It presents a
very useful task for product evaluation, social recommendation and popularity analysis. It constitutes also a crucial move towards
natural language processing (NLP) domains. Our main goal is to identify sentiments towards aspect in the sentence. An aspect
presents a specific entity or object features (price, product quality, etc.). Therefore, its analysis requires two primordial steps:
extract entity aspects and identify the sentiments from all these aspects.
In our research work, we propose a new hybrid method to detect sentiments towards aspects. We start with a machine learning
method to detect the different aspects within a given sentence, followed by a rule based method to identify sentiments within these
aspects. The evaluation of our hybrid system based on a reference dataset of Arabic Hotels’ reviews Semantic Evaluation-2016
shows that our system outperforms baseline research to achieve encouraging results (96% of F-score).
Keywords: Hybrid method, sentiment analysis, aspect detection, machine learning, rule-based method, Arabic language
1. Introduction 1
Sentiment analysis (SA) is considered as a classifi- 2
cation task. This classification can be binary to repre- 3
sent only positive and negative sentiment or multi-class 4
when the text presents the positive, negative and neu- 5
tral classes. SA can hold at various levels: document, 6
sentence and aspect. At the document level, the task is 7
to classify whether a whole opinionated document has 8
a positive, negative or neutral sentiment. At the sen- 9
tence level, the task is to classify whether an individ- 10
ual sentence has a positive, negative or neutral senti- 11
ment. At the aspect level, the task is to classify the sen- 12
timent of individual sentences or phrases intended to- 13
wards certain entities or aspects. When the SA is done 14
at aspect level, we speak about the Aspect Based Sen- 15
timent Analysis (ABSA). 16
Since this aspect is considered as the attributes of 17
an entity, the results in ABSA are more detailed, inter- 18
*
Corresponding author: Sana Trigui, Miracl, University of Sfax,
Sfax, Tunisia. E-mail: truguisana0@gmail.com.
esting and accurate. For example in the following sen- 19
tence: “The crew is excellent and the technique is bad”, 20
at the sentence level, the sentiment expressed is neu- 21
tral (positive and negative at the same time). Whereas, 22
at aspect level, the sentiment is positive for “the crew” 23
and negative for “the technique”. ABSA helps us to 24
decide whether specific item or service is good/bad or 25
preferred or not preferred. It is also useful to identify 26
opinions of people about any entity. Here, an entity can 27
be a product, person, event, organization, or topic on 28
which an opinion is expressed. This entity is composed 29
both of components and a set of attributes. For exam- 30
ple, a cell phone is composed of a screen, a battery, and 31
the attributes are the size and the weight. For the sake 32
of simplicity, components and attributes are referred to 33
as aspects. 34
Several studies on ABSA have already been per- 35
formed in many languages, such as English and French. 36
There is relatively less work on Arabic language [1] de- 37
spite Arabic is currently ranked as the fourth language 38
used in the web, and there are about 168 million of 39
1448-5869/20/$35.00 © 2020 – IOS Press. All rights reserved
uncorrected proof version