Vol 6, 2019

  1. EVALUATING THE IMPACT OF REMOVING LESS IMPORTANT TERMS ON SENTIMENT ANALYSIS  (page : 1 – 12)

 

Salhana Amad Darwis, Duc Nghia Pham, Ang Jia Pheng, Ong Hong Hoe

Artificial Intelligence Laboratory,

MIMOS Berhad

{salhana.darwis, nghia.pham,  jp.ang, hh.ong} @mimos.my

 

ABSTRACT

Sentiment analysis is an important task in Natural Language Processing (NLP) that analyses and predicts people’s opinion from textual data. It is a complex process due to the interactions with computer science, linguistics, psychology and social science disciplines. There is no straight forward rule to analyse and predict sentiment. Supervised learning methods, which adopt learning models from human, are being widely used by NLP researchers and experts to predict sentiment. However, this approach is tricky due to the challenges in ensuring the quality of the manually labelled training dataset. In this study, we investigated the use of linguistic factors to improve the model’s accuracy. We gathered two datasets: (i) 125,000 annotated sentences from Amazon product reviews, and (ii) 11,250 annotated sentences from financial news articles. We then pre-processed the data, identified the less important terms that exist in the dataset, the linguistic features and their effect towards the correctness of predicted sentiment. Our experimental results showed that punctuation separation and removal of supporting POS words improves precision accuracy in larger-generic dataset rather than in smaller-context sensitive dataset.

 Field of Research: sentiment analysis, supervised learning, NLP, linguistics

[Download PDF]

 

2. BENCHMARKING SUPERVISED LEARNING MODELS FOR SENTIMENT ANALYSIS 

(page : 13 – 30)

 

Ang Jia Pheng, Duc Nghia Pham, Ong Hong Hoe
Artificial Intelligence Laboratory MIMOS Berhad
{jp.ang, nghia.pham, hh.ong} @mimos.my

ABSTRACT

Sentiment analysis is an effective method to extract meaningful information from sentence and documents. While pursuing performance of a model, we also need to take other factors such as time, hardware and manpower into consideration. Beside algorithm, the characteristic and feature of textual dataset also affect the performance of a model. In this paper, we benchmark the performance of 1 deep learning model (Bidirectional long short-term memory – BiLSTM), and 1 machine learning model (FastText) using 3 different datasets: binary class document level Amazon Product Reviews dataset, binary class sentence level Sentiment140 Twitter Sentiment dataset, and ternary class sentence level financial domain-specific dataset. Since FastText does not support GPU for computation, this study ran on CPU to benchmark the speed and performance. Our results show that BiLSTM performed better than FastText in all tasks, with the cost of at least 15000% longer time for training process under the same machine specification.

Field of Research: sentiment analysis, machine learning, natural language processing.

[Download PDF]

 

3. BENCHMARKING SUPERVISED LEARNING MODELS FOR EMOTION ANALYSIS   (page : 31 – 38)

 

Ang Jia Pheng, Duc Nghia Pham, Ong Hong Hoe
Artificial Intelligence Laboratory MIMOS Berhad
{jp.ang, nghia.pham, hh.ong} @mimos.my

ABSTRACT

Emotion is the most genuine reaction of a person towards a circumstance or an object that are usually hidden between lines in their speech, text and actions. While emotion is more sophisticated and complicated to process and analyze, emotion provides more detailed and valuable insights for organizations to re-evaluate/fine-tune their actions and make informed decisions. This paper benchmarked supervised learning models for emotion analysis using sentence-level documents with six categories: anger, sadness, joy, love, surprise, and fear. We evaluated one deep learning model: Bidirectional Long-Short Term Memory (BiLSTM), and one machine learning model: FastText. Since FastText does not support GPU, both BiLSTM and FastText were ran on CPU for a fair time comparison. The results showed that while sacrificing speed that took at least 9000% longer to train and validate, BiLSTM consistently outperformed FastText. We also found that text pre-processing helped boost the performance of supervised learning models in emotion analysis.

Field of Research: emotion analysis, machine learning, natural language processing

[Download PDF]

 

4. DECISION SUPPORT SYSTEM TOWARDS INDUSTRIAL REVOLUTION 4.0

  (page : 39 – 56)

 

Syarbaini Ahmad
syarbaini@kuis.edu.my

Dept. of Comp Science
Faculty of Science and IT
Kolej Universiti Islam Antarabangsa Selangor
Bandar Sri Putra Kajang
Selangor, MALAYSIA

Nur Aisya Insyirah Manaf
aisyamnf@gmail.com

Jahirah Juwairiyah Abdul Latif
jahirahjuwairiyah@gmail.com

Siti Fairuz Mohd Nizam
fairuznizm@gmail.com

Dept. of Multimedia
Faculty of Science and IT
Kolej Universiti Islam Antarabangsa Selangor
Bandar Sri Putra Kajang
Selangor, MALAYSIA

ABSTRACT

Industrial Revolution (IR) 4.0 is not only the revolution of industry. It is about the innovation of the technology to move into a new paradigm and expansion of knowledge. This paper discussed the trend of decision support system (DSS) since the first industrial revolution. The trend of revolution is evolved in diverse areas to comply with the needs of human throughout the time. The early points of discussion are divided into three phases of industrial revolution which is industrial revolution one, two and three. From the previous IR we bring the flow to what will happen in the current and next industrial revolution. The aim of this paper is to discuss on how invention in past revolution help to support nowadays technology specially to assist human DSS. Trend of future technology are discussed at the end of this paper to see the enhancement of DSS in human life.

Field of Research: Decision Support System, industrial revolution, ir4.0, innovation, application

[Download PDF]

 

5. PENERIMAAN GOOGLE CLASSROOM DALAM MATA PELAJARAN SEJARAH TINGKATAN LIMA (page : 57 – 69)

Nor Zanira Binti Abd Manan
Fakulti Seni, Komputeran dan Industri Kreatif
Universiti Pendidikan Sultan Idris
zaniramanan@gmail.com

Hafizul Fahri Bin Hanafi
Fakulti Seni, Komputeran dan Industri Kreatif
Universiti Pendidikan Sultan Idris
hafizul@fskik.upsi.edu.my

ABSTRAK

Penyelidikan ini dijalankan bertujuan untuk melihat penerimaan pelajar tingkatan lima terhadap penggunaan Google Classroom bagi mata pelajaran Sejarah. Penyelidik menggunakan kaedah kajian tinjauan dengan menjadikan 40 pelajar tingkatan lima dari SMK Alam Megah, Shah Alam sebagai responden. Manakala kajian rintis pula telah dijalankan di SMK Seksyen 7 yang melibatkan 35 orang responden dan memperolehi alpha cronbach sebanyak 0.880. Data yang diperolehi telah dianalisis menggunakan analisis deskriptif melalui penggunaan perisian SPSS versi 23. Soal selidik bagi kajian ini diukur berdasarkan penggunaan model UTAUT (Unified Theory of Acceptance and Use of Technology). Hasil dapatan kajian menunjukkan semua pelajar berminat dan bersedia menggunakan Google Classroom dalam sesi pembelajaran mereka dan sekaligus ia membuktikan bahawa penggunaan Google Classroom sebagai BBM dapat menjadikan proses PdP lebih efektif dan bermakna. Dapatan kajian juga menunjukkan bahawa pengintegrasian Google Classroom sebagai BBM dalam PdP dapat diterima dengan baik oleh para pelajar dan seharusnya ia diteruskan oleh guru-guru bagi semua mata pelajaran.

Bidang Penyelidikan : Google Classroom, UTAUT, Pengajaran dan Pembelajaran (PdP), BBM (Bahan Bantu Mengajar)

[Download PDF]