DOES GOOGLE TRANSLATE AFFECT LEXICON-BASED SENTIMENT ANALYSIS OF MALAY SOCIAL MEDIA TEXT?
Vanessa Enjop1, Rosanita Adnan2∗, Nursuriati Jamil3, Sanizah Ahmad4, Zarina Zainol5 and Siti Arpah Ahmad6
National Autism Resource Centre, Faculty of Computer and Mathematical Sciences, Universiti Teknologi MARA, 40450 Shah Alam Selangor, Malaysia
1This email address is being protected from spambots. You need JavaScript enabled to view it., 2*This email address is being protected from spambots. You need JavaScript enabled to view it., 3This email address is being protected from spambots. You need JavaScript enabled to view it., 4This email address is being protected from spambots. You need JavaScript enabled to view it., 5This email address is being protected from spambots. You need JavaScript enabled to view it. and 6This email address is being protected from spambots. You need JavaScript enabled to view it.,
ABSTRACT
There are a lot of sentiment resources for English, however, there are limited resources in a resource-poor language like the Malay language. One approach to improving sentiment analysis is to translate the focus-language text to a resource-rich language such as English by using Machine Translation (MT). However, when text is translated from one language into another, sentiment is preserved to varying degrees. The objective of this paper is to assess the performance of MT in Google Translate towards sentiment analysis of Malay social media text on Facebook pages of a caregiver of a person with autism. A total of 3,525 Facebook comments in the Malay language were gathered from May to October 2020. The comments were manually translated to English to create dataset_manual. Google Translate was used to automatically translate the Malay comments into English creating dataset_auto. The sentiment polarity of each comment was labeled as a ground truth dataset. A lexicon-based approach was used to extract sentiment from both dataset_manual and dataset_auto to determine the sentiment polarity. Results show that 65.9% of sentiment analysis using dataset_auto significantly reduces sentiment analysis. The sentiment expressions are often mistranslated into neutral expressions when translated. Meanwhile, sentiment analysis using dataset_manual was still able to capture the sentiment of Facebook comment without taking the comment out of context where 92.5% shows positive sentiment towards comments related to autism spectrum disorder.
Keywords: Facebook, Google Translate, Lexicon-Based, Machine Translation, Sentiment Analysis
Published On: 1 October 2022
