Dani Gunawan
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Dani Gunawan has a total of 78 co-authors. Here are the top co-authors by number of shared publications:
- Romi Fadillah Rahmat
Universitas Sumatera Utara
14 shared publications Amalia Amalia
Department of Computer Science
8 shared publicationsA Amalia
7 shared publications- Amalia Amalia
Universitas Sumatera Utara
4 shared publications - Baihaqi Siregar
Department of Information Technology, Universitas Sumatera Utara, Medan, Indonesia
3 shared publications - Fahmi Fahmi
Universitas Sumatera Utara
3 shared publications
Latest Publications
- Open Access
- 26 May 2026
- Published by PeerJ in PeerJ Computer Science
- Vol. 12, e3927
Emoji-aware multimodal representations for multi-label emotion analysis in Indonesian social media
Text Emotion Analysis (TEA) has progressed rapidly, but many studies still approach emotion as a single-label problem and focus solely on text. Research on TEA in low-resource languages such as Indonesian remains relatively scarce, and most available datasets do not explicitly capture paralinguistic cues, such as emojis, even though emojis often convey strong emotional signals. To bridge this gap, we propose an emoji-aware multimodal framework for multi-label emotion classification in Indonesian social media, using the eight Plutchik emotions as labels. We constructed the dataset using a semi-automated annotation pipeline: distant supervision was used to bootstrap labels, and experts then refined them by considering both text and emoji signals. The release contains 139,414 cleaned Indonesian social media posts from YouTube, X (formerly Twitter), Reddit, and Steam. We provide (i) a mixed corpus with posts both with and without emojis and (ii) an emoji-rich subset for emoji-aware experiments, along with the annotation guidelines and fixed train/validation/test splits. For all model variants, Indonesian Bidirectional Encoder Representations from Transformers (IndoBERT) serves as the text backbone; the only difference lies in how emoji information is incorporated. We tested three integration options for emojis: adding them early (by concatenation or tokenizer expansion), fusing them mid-stream with cross-attention, or combining predictions late (average/weighted). Compared to a text-only baseline, emoji-aware models perform better overall, reflecting the fact that emojis often convey emotional cues that text alone misses. The best scores come from tokenizer expansion (micro-averaged F1 = 0.8554; macro-averaged F1 = 0.7769), likely because emojis are processed alongside words through IndoBERT’s contextual layers. Beyond model comparisons, we also introduce an Indonesian emoji-aware multi-label dataset and release fixed splits and guidelines to support reproducible research.
- 26 May 2026
- Published by PeerJ in PeerJ Computer Science
- Vol. 12, e3927
Open AccessEmoji-aware multimodal representations for multi-label emotion analysis in Indonesian social media
Text Emotion Analysis (TEA) has progressed rapidly, but many studies still approach emotion as a single-label problem and focus solely on text. Research on TEA in low-resource languages such as Indonesian remains relatively scarce, and most available datasets do not explicitly capture paralinguistic cues, such as emojis, even though emojis often convey strong emotional signals. To bridge this gap, we propose an emoji-aware multimodal framework for multi-label emotion classification in Indonesian social media, using the eight Plutchik emotions as labels. We constructed the dataset using a semi-automated annotation pipeline: distant supervision was used to bootstrap labels, and experts then refined them by considering both text and emoji signals. The release contains 139,414 cleaned Indonesian social media posts from YouTube, X (formerly Twitter), Reddit, and Steam. We provide (i) a mixed corpus with posts both with and without emojis and (ii) an emoji-rich subset for emoji-aware experiments, along with the annotation guidelines and fixed train/validation/test splits. For all model variants, Indonesian Bidirectional Encoder Representations from Transformers (IndoBERT) serves as the text backbone; the only difference lies in how emoji information is incorporated. We tested three integration options for emojis: adding them early (by concatenation or tokenizer expansion), fusing them mid-stream with cross-attention, or combining predictions late (average/weighted). Compared to a text-only baseline, emoji-aware models perform better overall, reflecting the fact that emojis often convey emotional cues that text alone misses. The best scores come from tokenizer expansion (micro-averaged F1 = 0.8554; macro-averaged F1 = 0.7769), likely because emojis are processed alongside words through IndoBERT’s contextual layers. Beyond model comparisons, we also introduce an Indonesian emoji-aware multi-label dataset and release fixed splits and guidelines to support reproducible research.
- Open Access
- 26 October 2025
- Published by Association for Computing Machinery (ACM)
- p. 11-19
Efficient and Accurate Post-Training Sparsification of Large Language Models with Proximal Operators
- 26 October 2025
- Published by Association for Computing Machinery (ACM)
- p. 11-19
Open AccessEfficient and Accurate Post-Training Sparsification of Large Language Models with Proximal Operators
- 11 November 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 8, 106-110
News Articles Classification for Electronic Information and Transaction Law in Indonesia using Support Vector Machine
Currently, in Indonesia there are many cases related to the Electronic Information and Transaction (ITE) Law. The ITE Law is a law that regulates violations in the digital world and the internet. Most of the people in Indonesia do not have the correct understanding of ITE law, and even often misinterpret the existing verses. Meanwhile, the media in Indonesia often reports about violations of the ITE Law in their printing and digital media. However, due to the lack of good socialization, people's understanding does not increase, which also effect the increasing ITE cases occurred. To see the violations of these cases in more depth, we should see the official crime document provided by law enforcement elements. Because it is necessary to provide early education to the public, both adults and adolescents, regarding the violations of ITE Law, it is important to build a system that is able to classify news articles according to the violations committed automatically and quickly. Grouping news articles can use one of the classification algorithms, namely Support Vector Machine (SVM). This study utilizes SVM in classifying 5 categories related to violations of the ITE Law in Indonesia. This research produces a classification system with RBF kernel with an accuracy of 86.27%.
- 11 November 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 8, 106-110
News Articles Classification for Electronic Information and Transaction Law in Indonesia using Support Vector Machine
Currently, in Indonesia there are many cases related to the Electronic Information and Transaction (ITE) Law. The ITE Law is a law that regulates violations in the digital world and the internet. Most of the people in Indonesia do not have the correct understanding of ITE law, and even often misinterpret the existing verses. Meanwhile, the media in Indonesia often reports about violations of the ITE Law in their printing and digital media. However, due to the lack of good socialization, people's understanding does not increase, which also effect the increasing ITE cases occurred. To see the violations of these cases in more depth, we should see the official crime document provided by law enforcement elements. Because it is necessary to provide early education to the public, both adults and adolescents, regarding the violations of ITE Law, it is important to build a system that is able to classify news articles according to the violations committed automatically and quickly. Grouping news articles can use one of the classification algorithms, namely Support Vector Machine (SVM). This study utilizes SVM in classifying 5 categories related to violations of the ITE Law in Indonesia. This research produces a classification system with RBF kernel with an accuracy of 86.27%.
- Open Access
- 1 May 2020
- Published by IOP Publishing in IOP Conference Series: Materials Science and Engineering
- Vol. 851  (1)
The Development of Mobile and Server Application for Traffic Data Collection Based on Location-Based Services (LBS)
Traffic data includes various data such as the number of vehicles, vehicle velocity, or travel time that might lead to meaningful information for the traffic engineers. This research aims to collect the traffic data based on the Location Based Services (LBS) by utilizing the Global Positioning System (GPS) in a smartphone. The drivers bring the smartphone and collecting the traffic data while driving the vehicle. It is done in real-time by running the mobile application in background service. The data includes the date, timestamp, latitude, longitude, device ID, and vehicle velocity. These data are temporarily stored in the internal smartphone database, then synchronized to the database in the server application. According to the result, the large number of online taxi drivers and a longer time to collect the data should be considered to cover a wide area.
- 1 May 2020
- Published by IOP Publishing in IOP Conference Series: Materials Science and Engineering
- Vol. 851  (1)
Open AccessThe Development of Mobile and Server Application for Traffic Data Collection Based on Location-Based Services (LBS)
Traffic data includes various data such as the number of vehicles, vehicle velocity, or travel time that might lead to meaningful information for the traffic engineers. This research aims to collect the traffic data based on the Location Based Services (LBS) by utilizing the Global Positioning System (GPS) in a smartphone. The drivers bring the smartphone and collecting the traffic data while driving the vehicle. It is done in real-time by running the mobile application in background service. The data includes the date, timestamp, latitude, longitude, device ID, and vehicle velocity. These data are temporarily stored in the internal smartphone database, then synchronized to the database in the server application. According to the result, the large number of online taxi drivers and a longer time to collect the data should be considered to cover a wide area.
- Open Access
- 1 October 2019
- Published by IOP Publishing in IOP Conference Series: Materials Science and Engineering
- Vol. 648  (1) , 012011
Automatic Number Plate Recognition for Indonesian License Plate by Using K-Nearest Neighbor Algorithm
Transportation surveillance requires many human resources to cover all the road in a city. The human resources might be reduced by utilizing automation. For example, a task to recognize the license plate can be automated by utilizing CCTVs or surveillance cameras that usually installed in the crossroads. The automatic number plate recognition is required to record the license plate of the vehicle that violates the traffic rules. This number can be processed for future references. This research utilizes K-Nearest Neighbor (KNN) to recognize the license plate. The research is conducted to 125 license plate images which are divided into 100 training images and 25 testing images. The success rate of this research is 92.86%. The condition of the license plate and light intensity influence the recognition result.
- 1 October 2019
- Published by IOP Publishing in IOP Conference Series: Materials Science and Engineering
- Vol. 648  (1) , 012011
Open AccessAutomatic Number Plate Recognition for Indonesian License Plate by Using K-Nearest Neighbor Algorithm
Transportation surveillance requires many human resources to cover all the road in a city. The human resources might be reduced by utilizing automation. For example, a task to recognize the license plate can be automated by utilizing CCTVs or surveillance cameras that usually installed in the crossroads. The automatic number plate recognition is required to record the license plate of the vehicle that violates the traffic rules. This number can be processed for future references. This research utilizes K-Nearest Neighbor (KNN) to recognize the license plate. The research is conducted to 125 license plate images which are divided into 100 training images and 25 testing images. The success rate of this research is 92.86%. The condition of the license plate and light intensity influence the recognition result.
- Open Access
- 1 June 2019
- Published by IOP Publishing in Journal of Physics: Conference Series
- Vol. 1235  (1)
Automated Bahasa Indonesia essay evaluation with latent semantic analysis
Essay Examination is one of many ways to evaluate the learning process of students. The Essay test can measure a student’s capability of memory and to develop an idea. With the vast development of Information Technology, an essay test is done in a more sophisticated way with a platform called e-learning. This research implemented an e-learning interface to conduct an essay test, starting with giving out questions, answering it and marking it. The system will generate an automatic mark, using the Latent Semantic Analysis (LSA) method by measuring the relevance of the key answer and the student’s answer. Before entering the LSA phase, the answer will go through pre-processing step which is cleansing, case folding, tokenization, stop word, convert negation and stemming. The result of the accuracy of this system compared to teacher’s manual assessment is 83.3%. Expectedly, this research will help teachers in making their assessment process more efficient.
- 1 June 2019
- Published by IOP Publishing in Journal of Physics: Conference Series
- Vol. 1235  (1)
Open AccessAutomated Bahasa Indonesia essay evaluation with latent semantic analysis
Essay Examination is one of many ways to evaluate the learning process of students. The Essay test can measure a student’s capability of memory and to develop an idea. With the vast development of Information Technology, an essay test is done in a more sophisticated way with a platform called e-learning. This research implemented an e-learning interface to conduct an essay test, starting with giving out questions, answering it and marking it. The system will generate an automatic mark, using the Latent Semantic Analysis (LSA) method by measuring the relevance of the key answer and the student’s answer. Before entering the LSA phase, the answer will go through pre-processing step which is cleansing, case folding, tokenization, stop word, convert negation and stemming. The result of the accuracy of this system compared to teacher’s manual assessment is 83.3%. Expectedly, this research will help teachers in making their assessment process more efficient.
- Open Access
- 1 June 2019
- Published by IOP Publishing in Journal of Physics: Conference Series
- Vol. 1235  (1)
Multi-language Dictionary Integration by Using Semantic Web
- 1 June 2019
- Published by IOP Publishing in Journal of Physics: Conference Series
- Vol. 1235  (1)
Open AccessMulti-language Dictionary Integration by Using Semantic Web
- Open Access
- 1 June 2019
- Published by IOP Publishing in Journal of Physics: Conference Series
- Vol. 1235  (1)
Building the Application to Identify Incorrect Capital Letters Writing in Bahasa Indonesia
Ejaan Bahasa Indonesia (EBI) is a set of writing rules in Bahasa Indonesia which is implemented since 2015 according to the government rule Peraturan Kementerian Pendidikan dan Kebudayaan Republik Indonesia Nomor 5 Tahun 2015. This new set of rules replaces previously implemented rules which are known as Ejaan Yang Disempurnakan (EYD). The capital letter rule is one of the elements of EBI. According to the official dictionary of Bahasa Indonesia, the capital letter is a letter which has a bigger size and particular form than the common letter. This research yields an application to identify the incorrect capital letter writing in Bahasa Indonesia based on the user input. The application implements Part of Speech Tagging (POS), Named Entity Recognition (NER), N-Gram and set of rules based on EBI. As the conclusion, the application is able to check 81.57% words that should be written in capital form.
- 1 June 2019
- Published by IOP Publishing in Journal of Physics: Conference Series
- Vol. 1235  (1)
Open AccessBuilding the Application to Identify Incorrect Capital Letters Writing in Bahasa Indonesia
Ejaan Bahasa Indonesia (EBI) is a set of writing rules in Bahasa Indonesia which is implemented since 2015 according to the government rule Peraturan Kementerian Pendidikan dan Kebudayaan Republik Indonesia Nomor 5 Tahun 2015. This new set of rules replaces previously implemented rules which are known as Ejaan Yang Disempurnakan (EYD). The capital letter rule is one of the elements of EBI. According to the official dictionary of Bahasa Indonesia, the capital letter is a letter which has a bigger size and particular form than the common letter. This research yields an application to identify the incorrect capital letter writing in Bahasa Indonesia based on the user input. The application implements Part of Speech Tagging (POS), Named Entity Recognition (NER), N-Gram and set of rules based on EBI. As the conclusion, the application is able to check 81.57% words that should be written in capital form.
- 1 November 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-6
Filtering Spam Text Messages by Using Twitter-LDA Algorithm
Recently, the usage of short messaging service (SMS) or text messages have been changed gradually to product or service promotion, and even fraud. The mobile phone users in Indonesia also experience the same condition. A simple approach to address this issue is creating black list of phone numbers or certain keywords and phrases. However, this approach is inefficient because the spammer might change the phone number or change the content of the text message. Meanwhile, another approach is utilizing text classification such as Naive Bayes, k-Nearest Neighbor (kNN), and Support Vector Machine (SVM) to recognize pattern of the text messages. This research proposes Twitter-LDA algorithm to identify spam text messages in Bahasa Indonesia. There are total 985 text messages divided to 774 text messages for training dataset and 211 text messages for testing dataset. These datasets consist of 860 spam and 125 ham text messages. All the text messages should be pre-processed before the training and testing process are applied. This research conducts five experiments which yield the average of f-score is 94.26% and accuracy is 96.49%. According to this result, the Twitter-LDA algorithm has demonstrated a good performance in identifying spam text messages in Bahasa Indonesia.
- 1 November 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-6
Filtering Spam Text Messages by Using Twitter-LDA Algorithm
Recently, the usage of short messaging service (SMS) or text messages have been changed gradually to product or service promotion, and even fraud. The mobile phone users in Indonesia also experience the same condition. A simple approach to address this issue is creating black list of phone numbers or certain keywords and phrases. However, this approach is inefficient because the spammer might change the phone number or change the content of the text message. Meanwhile, another approach is utilizing text classification such as Naive Bayes, k-Nearest Neighbor (kNN), and Support Vector Machine (SVM) to recognize pattern of the text messages. This research proposes Twitter-LDA algorithm to identify spam text messages in Bahasa Indonesia. There are total 985 text messages divided to 774 text messages for training dataset and 211 text messages for testing dataset. These datasets consist of 860 spam and 125 ham text messages. All the text messages should be pre-processed before the training and testing process are applied. This research conducts five experiments which yield the average of f-score is 94.26% and accuracy is 96.49%. According to this result, the Twitter-LDA algorithm has demonstrated a good performance in identifying spam text messages in Bahasa Indonesia.
- 1 October 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-6
Early Identification of Leaf Stain Disease in Sugar Cane Plants Using Speeded-Up Method Robust Features
Cane (Saccharum officinarumL.) is a plant that is used as the main raw material in the sugar industry. One of the factors that can cause the reduction of sugar cane production and inhibit the development of the sugar industry is the late-diagnosed of leaf stain disease in sugarcane crops that cause crop failure. It is quite often the case because of the limited number of service facilities for communication and also the high cost to be incurred if the sugar cane entrepreneurs want to conduct the inspection process of the crop. Therefore, this study has a goal to identify leaf stain disease in sugarcane plants. This research used the image of sugar cane leaves that have stain spots as input for the image processing process using the Speeded-Up Robust Features (SURF) method. The identification process of leaf stains is divided into three types of leaf stain disease, namely ring stains, rust stains, and yellow stains. In this study, 60 images were used as preliminary data and 30 images for the testing process. Following the conducted test, it can be concluded that the early identification process of leaf stain disease in sugar cane has an accuracy rate of 90%.
- 1 October 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-6
Early Identification of Leaf Stain Disease in Sugar Cane Plants Using Speeded-Up Method Robust Features
Cane (Saccharum officinarumL.) is a plant that is used as the main raw material in the sugar industry. One of the factors that can cause the reduction of sugar cane production and inhibit the development of the sugar industry is the late-diagnosed of leaf stain disease in sugarcane crops that cause crop failure. It is quite often the case because of the limited number of service facilities for communication and also the high cost to be incurred if the sugar cane entrepreneurs want to conduct the inspection process of the crop. Therefore, this study has a goal to identify leaf stain disease in sugarcane plants. This research used the image of sugar cane leaves that have stain spots as input for the image processing process using the Speeded-Up Robust Features (SURF) method. The identification process of leaf stains is divided into three types of leaf stain disease, namely ring stains, rust stains, and yellow stains. In this study, 60 images were used as preliminary data and 30 images for the testing process. Following the conducted test, it can be concluded that the early identification process of leaf stain disease in sugar cane has an accuracy rate of 90%.