Page 141 - The-5th-MCAIT2021-eProceeding

Analyzing Twitter Reviews on Halal Food using Sentiment
Analysis

a
b
Alya Nur Adlina Ahmad Nazri , Siti Nur Kamaliah Kamarudin
a,b Faculty of Computer and Mathematical Sciences, Universiti Teknologi MARA, Shah Alam, Selangor, Malaysia
* Email: alyaadlinazri@gmail.com; snkamaliah@uitm.edu.my

Abstract

Malaysia have among the highest number of social media users, and one of the popular topics discussed online is the halal
status of trendy restaurants or popular food. Since most users nowadays use social media like Twitter to express their
opinions, it is also convenient for them to search information about the halal status of their place or food of interest. However,
it can be time consuming for users to confirm the halal status of some places or food in Malaysia as some places may serve
halal food but they do not bother to get the Malaysian halal certification. Therefore, this research explores the process of
how to evaluate the sentiments on halal food or restaurants from Twitter tweets and to identify the halal, non-halal or pork-
free status of the food or restaurant. A sentiment analysis was performed using a Lexicon-based approach to predict and
label the tweets. Subsequently, Machine Learning classifiers such as Support Vector Machine (SVM), Naïve Bayes, and
Deep Learning were applied to compare the sentiment analysis performance. The total number of tweets used in this research
was 71,333 tweets where as a result, Deep Learning method achieved the highest accuracy with 73.18% using SentiWordNet
on N-grams range of [1,3]. From these findings, Deep Learning outperformed other classifiers as it scores the highest for
both accuracy and f1-score metrics.

Keywords: halal food; Twitter reviews; sentiment analysis; machine learning; lexicon-based approach

1. Introduction
The halal industry are synonymous with areas like tourism, cosmetics and of course, food. Other industries
such as pharmaceuticals and healthcare goods have also started to apply the concept of halal (Sulaiman et al.
2018) due to the increase of demand for halal items. Twitter has been one of the most popular social media
platforms to be used for sentiment analysis (Tyagi & Tripathi 2019). It’s functionalities is similar to a huge
forum where users can contribute their ideas, thoughts, and opinions making it an ideal platform to analyze
different opinions on various areas (Sarlan et al. 2015). In 2019, around 2.5 million Malaysians were active
users on Twitter (Malaysia: number of Twitter users 2014-2019 2019) where the users posted, tweeted and
shared many opinions on various issues. One of the commonly debated issues on Twitter is on the halal-ness of
a certain food or restaurant, especially if the food or restaurant suddenly became popular in the country
(Feizollah et al. 2019).
The huge number of Malaysian Twitter users expressing their personal opinions via twitter benefits those
who would like to search or read reviews on certain food or restaurant. However, it can also be time-consuming
to go through all of the reviews. Additionally, some tweets may use short forms or unfamiliar terms which
makes it even more time-consuming for users to sift through the tweets to find the halal status for a particular
food or restaurant. Therefore, for this research the author chose to investigate and classify tweets on halal food
reviews taken from Twitter in order to help users search for the halal status of a particular food or restaurant.
Subsequently, this research attempted to identify some of the keywords used to determine the halal status of the
food in the tweets by users. This research also make use of several machine learning approaches to classify the
tweets and finally, results will be displayed through a dashboard which was incorporated into Halalopedia, a
web-based system created using the results from this research.

E- Proceedings of The 5th International Multi-Conference on Artificial Intelligence Technology (MCAIT 2021) [128]
Artificial Intelligence in the 4th Industrial Revolution