Pub Date : 2022-12-17DOI: 10.1109/ICCIT57492.2022.10054672
A. Shafi, Md. Mareful Hasan Maruf, Sunanda Das
Pneumonia is said to be the "Silent Killer" disease caused by the infection of virus, bacteria, or fungi in the lung alveoli. It bears an extensive risk for people, especially children in some developing nations. The ecumenic way to detect pneumonia is from Chest X-ray data. But it has some complications to diagnose pneumonia if the lung has gone through some surgery, bleeding, the superabundance of fluids, or lung cancer. So, it is necessary to take the help of Computer-Aided Diagnosis (CAD) which can collaborate the doctors to detect pneumonia. Many deep learning methods are applicable to detect pneumonia. Our research introduces a new model generated from the fusion of two different transfer learning models, the Xception model and the VGG16 model. Our research includes image pre-processing using image normalization and augmentation. We took two different transfer learning models namely Xception, and VGG16 for the feature extraction, then added some layers, made a fusion, and lastly added some extra dense layers to develop the proposed model. We took 5216 images of two classes named ‘NORMAL’ and ‘PNEUMONIA’ images to train our model. We took 5216 images to train the model in ‘NORMAL’ and ‘PNEUMONIA’ form. The results were tested with 624 images belonging to two classes. The proposed model achieved accuracy, precision, recall, and f1-score of 91.67%, 92.30%, 89.92%, and 90.87% respectively. The extensive experimental analysis demonstrates the viability of the proposed approach for various test samples.
{"title":"Pneumonia Detection from Chest X-ray Images Using Transfer Learning by Fusing the Features of Pre-trained Xception and VGG16 Networks","authors":"A. Shafi, Md. Mareful Hasan Maruf, Sunanda Das","doi":"10.1109/ICCIT57492.2022.10054672","DOIUrl":"https://doi.org/10.1109/ICCIT57492.2022.10054672","url":null,"abstract":"Pneumonia is said to be the \"Silent Killer\" disease caused by the infection of virus, bacteria, or fungi in the lung alveoli. It bears an extensive risk for people, especially children in some developing nations. The ecumenic way to detect pneumonia is from Chest X-ray data. But it has some complications to diagnose pneumonia if the lung has gone through some surgery, bleeding, the superabundance of fluids, or lung cancer. So, it is necessary to take the help of Computer-Aided Diagnosis (CAD) which can collaborate the doctors to detect pneumonia. Many deep learning methods are applicable to detect pneumonia. Our research introduces a new model generated from the fusion of two different transfer learning models, the Xception model and the VGG16 model. Our research includes image pre-processing using image normalization and augmentation. We took two different transfer learning models namely Xception, and VGG16 for the feature extraction, then added some layers, made a fusion, and lastly added some extra dense layers to develop the proposed model. We took 5216 images of two classes named ‘NORMAL’ and ‘PNEUMONIA’ images to train our model. We took 5216 images to train the model in ‘NORMAL’ and ‘PNEUMONIA’ form. The results were tested with 624 images belonging to two classes. The proposed model achieved accuracy, precision, recall, and f1-score of 91.67%, 92.30%, 89.92%, and 90.87% respectively. The extensive experimental analysis demonstrates the viability of the proposed approach for various test samples.","PeriodicalId":255498,"journal":{"name":"2022 25th International Conference on Computer and Information Technology (ICCIT)","volume":"20 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133809324","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-12-17DOI: 10.1109/ICCIT57492.2022.10055451
Puja Das, Rupak Sharma, Sourav Dey Roy, N. Nath, M. Bhowmik
One of the most occurred cancers which cause death in women is breast cancer, contributing to 16% of all female cancers worldwide. Detection of the disease in the preliminary stage is the only way to treat the disease from any severity. Presence of the digital imaging modalities also allows the computerized diagnosis of a disease which overcomes the limitations of a human perception system. However, a pathologist who is knowledgeable and skilled is necessary for an appropriate diagnosis. Also, tissue sample analysis requires a lot of manual labor. Therefore, combining digital histopathology with computer-aided diagnostic (CAD) tools can assist in solving these issues. In this paper, we have proposed a hybrid framework of nucleus region segmentation from the histopathological images. The primary aim of the proposed framework is to ensemble information from multiple segmentations and, finally, fuse this information (in terms of intersection) to acquire the core and stable nucleus region(s). For this, we have ensemble the U-net model (with VGG-16 as the backbone network) with the fuzzy c-means algorithm for precise nucleus regions segmentation from the histopathological images. Experimental results reveal that the proposed framework performed better for cell nucleus segmentation with dice similarity index values of 0.8517 and 0.9357 using publicly available BreakHis and BreCaHAD, respectively.
{"title":"Ensemble Segmentation of Nucleus Regions from Histopathological Images towards Breast Abnormality Detection","authors":"Puja Das, Rupak Sharma, Sourav Dey Roy, N. Nath, M. Bhowmik","doi":"10.1109/ICCIT57492.2022.10055451","DOIUrl":"https://doi.org/10.1109/ICCIT57492.2022.10055451","url":null,"abstract":"One of the most occurred cancers which cause death in women is breast cancer, contributing to 16% of all female cancers worldwide. Detection of the disease in the preliminary stage is the only way to treat the disease from any severity. Presence of the digital imaging modalities also allows the computerized diagnosis of a disease which overcomes the limitations of a human perception system. However, a pathologist who is knowledgeable and skilled is necessary for an appropriate diagnosis. Also, tissue sample analysis requires a lot of manual labor. Therefore, combining digital histopathology with computer-aided diagnostic (CAD) tools can assist in solving these issues. In this paper, we have proposed a hybrid framework of nucleus region segmentation from the histopathological images. The primary aim of the proposed framework is to ensemble information from multiple segmentations and, finally, fuse this information (in terms of intersection) to acquire the core and stable nucleus region(s). For this, we have ensemble the U-net model (with VGG-16 as the backbone network) with the fuzzy c-means algorithm for precise nucleus regions segmentation from the histopathological images. Experimental results reveal that the proposed framework performed better for cell nucleus segmentation with dice similarity index values of 0.8517 and 0.9357 using publicly available BreakHis and BreCaHAD, respectively.","PeriodicalId":255498,"journal":{"name":"2022 25th International Conference on Computer and Information Technology (ICCIT)","volume":"51 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134327996","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-12-17DOI: 10.1109/ICCIT57492.2022.10055199
T. Hassan, Md. Munjure Mowla
Millimeter-wave (mmWave) technology is considered one of the major technologies for fifth-generation (5G) frameworks to meet the massive thirst for data traffic by guaranteeing enormous transmission capacity in the uplink, downlink, and backhaul links. However, medium access network (MAC) scheduling and admission control with respect to transmission control protocol (TCP) are still complicated in 5G heterogeneous networks. In addition, with the increasing number of users using mmWave communications, network attributes could be changed. In this paper, we design and implement a proportional fair (PF) scheduling algorithm in the MAC layer using network simulator-3 (ns3). This research inspects the downlink resource allocation among multiple users simultaneously in 5G heterogeneous networks. The simulation result depicts that the proposed approach outperforms the existing approach by an overall 14% in terms of SINR and throughput. The investigation through the proposed scheduler might be used to show the capability of 28 GHz frequency for mmWave communication and its commendable for future 5G systems with more complex structures.
{"title":"An Efficient Proportional Fair MAC Scheduling for Resource Allocation in 5G Millimeter Wave Networks","authors":"T. Hassan, Md. Munjure Mowla","doi":"10.1109/ICCIT57492.2022.10055199","DOIUrl":"https://doi.org/10.1109/ICCIT57492.2022.10055199","url":null,"abstract":"Millimeter-wave (mmWave) technology is considered one of the major technologies for fifth-generation (5G) frameworks to meet the massive thirst for data traffic by guaranteeing enormous transmission capacity in the uplink, downlink, and backhaul links. However, medium access network (MAC) scheduling and admission control with respect to transmission control protocol (TCP) are still complicated in 5G heterogeneous networks. In addition, with the increasing number of users using mmWave communications, network attributes could be changed. In this paper, we design and implement a proportional fair (PF) scheduling algorithm in the MAC layer using network simulator-3 (ns3). This research inspects the downlink resource allocation among multiple users simultaneously in 5G heterogeneous networks. The simulation result depicts that the proposed approach outperforms the existing approach by an overall 14% in terms of SINR and throughput. The investigation through the proposed scheduler might be used to show the capability of 28 GHz frequency for mmWave communication and its commendable for future 5G systems with more complex structures.","PeriodicalId":255498,"journal":{"name":"2022 25th International Conference on Computer and Information Technology (ICCIT)","volume":"33 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132212598","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-12-17DOI: 10.1109/ICCIT57492.2022.10055998
Nafisa Anjum Antora, Md. Ashiqur Rahman, A. Mosharraf, Mehrab Ibn Ehsan, M. Alve, M. M. Elahi
Waste management is a challenging task in this modern era and different approaches are still being discovered to make the separation of waste more efficient using the state-of-the-art technologies like Internet of Things (IoT), Edge-cloud integration, machine learning etc. In this work, we have designed and implemented an IoT-based smart bin that will use a machine learning algorithm to separate different types of wastes and send the data to the cloud server. It will sort organic and inorganic waste materials in an efficient manner using three different collection boxes built in. The wastes will be scanned through several sensors and image classification was used to separate the wastes in the designated collection boxes using the sensors. The sensor will also detect the level of the waste collected in the boxes and inform the personnel of the waste collection. The sensor will send a signal when it detects 70-80% filled boxes, which will give them enough time to get to the collection point. Each bin will have its own GPS tracking system to locate its location and also to avoid the hassle of being stolen. Although this smart bin will be battery powered, to make it eco-friendlier and economical, solar power will be used to recharge the battery. A working prototype has been developed as a proof-of-concept and preliminary results prove the efficiency of the proposed smart bin.
{"title":"Design and Implementation of a Smart Bin using IOT for an Efficient Waste Management System","authors":"Nafisa Anjum Antora, Md. Ashiqur Rahman, A. Mosharraf, Mehrab Ibn Ehsan, M. Alve, M. M. Elahi","doi":"10.1109/ICCIT57492.2022.10055998","DOIUrl":"https://doi.org/10.1109/ICCIT57492.2022.10055998","url":null,"abstract":"Waste management is a challenging task in this modern era and different approaches are still being discovered to make the separation of waste more efficient using the state-of-the-art technologies like Internet of Things (IoT), Edge-cloud integration, machine learning etc. In this work, we have designed and implemented an IoT-based smart bin that will use a machine learning algorithm to separate different types of wastes and send the data to the cloud server. It will sort organic and inorganic waste materials in an efficient manner using three different collection boxes built in. The wastes will be scanned through several sensors and image classification was used to separate the wastes in the designated collection boxes using the sensors. The sensor will also detect the level of the waste collected in the boxes and inform the personnel of the waste collection. The sensor will send a signal when it detects 70-80% filled boxes, which will give them enough time to get to the collection point. Each bin will have its own GPS tracking system to locate its location and also to avoid the hassle of being stolen. Although this smart bin will be battery powered, to make it eco-friendlier and economical, solar power will be used to recharge the battery. A working prototype has been developed as a proof-of-concept and preliminary results prove the efficiency of the proposed smart bin.","PeriodicalId":255498,"journal":{"name":"2022 25th International Conference on Computer and Information Technology (ICCIT)","volume":"27 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133630959","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-12-17DOI: 10.1109/ICCIT57492.2022.10055456
Azmain Yakin Srizon, S. Hasan, Md. Farukuzzaman Faruk, Abu Sayeed, Md. Ali Hossain
Throughout the last decades, human activity recognition has been considered one of the most complex tasks in the domain of computer vision. Previously, many works have suggested different machine learning models for the recognition of human actions from sensor-based data and video-based data which is not cost-efficient. The recent advancement of the convolutional neural network (CNN) has opened the possibility of accurate human activity recognition from still images. Although many researchers have already proposed some deep learning-based approaches addressing the problem, due to the high diversity in human actions, those approaches failed to achieve decent performance for all human actions under consideration. Some researchers argued that an ensemble of different models may work better in this regard. However, as the images used for recognition in this domain are mostly captured by security cameras, often, the deep models couldn’t extract valuable features resulting in misclassifications. To resolve these issues, in this study, we have considered three transfer-learned models i.e., DenseNet201, Xception, and EfficientNetB6, and applied a multichannel attention module to extract more distinguishable features. Moreover, a custom-made low-cost CNN has been proposed that works with small images extracting features that often get lost due to deep computations. Finally, the fusion of features extracted by attention-based transfer-learned models and the low-cost CNN has been used for the final prediction. We validated the proposed ensemble model on Stanford 40 actions, BU-101, and Willow datasets, and it achieved 97.48%, 98.29%, and 94.19% overall accuracy respectively which outperformed the previous performances by notable margins.
{"title":"Human Activity Recognition Utilizing Ensemble of Transfer-Learned Attention Networks and a Low-Cost Convolutional Neural Architecture","authors":"Azmain Yakin Srizon, S. Hasan, Md. Farukuzzaman Faruk, Abu Sayeed, Md. Ali Hossain","doi":"10.1109/ICCIT57492.2022.10055456","DOIUrl":"https://doi.org/10.1109/ICCIT57492.2022.10055456","url":null,"abstract":"Throughout the last decades, human activity recognition has been considered one of the most complex tasks in the domain of computer vision. Previously, many works have suggested different machine learning models for the recognition of human actions from sensor-based data and video-based data which is not cost-efficient. The recent advancement of the convolutional neural network (CNN) has opened the possibility of accurate human activity recognition from still images. Although many researchers have already proposed some deep learning-based approaches addressing the problem, due to the high diversity in human actions, those approaches failed to achieve decent performance for all human actions under consideration. Some researchers argued that an ensemble of different models may work better in this regard. However, as the images used for recognition in this domain are mostly captured by security cameras, often, the deep models couldn’t extract valuable features resulting in misclassifications. To resolve these issues, in this study, we have considered three transfer-learned models i.e., DenseNet201, Xception, and EfficientNetB6, and applied a multichannel attention module to extract more distinguishable features. Moreover, a custom-made low-cost CNN has been proposed that works with small images extracting features that often get lost due to deep computations. Finally, the fusion of features extracted by attention-based transfer-learned models and the low-cost CNN has been used for the final prediction. We validated the proposed ensemble model on Stanford 40 actions, BU-101, and Willow datasets, and it achieved 97.48%, 98.29%, and 94.19% overall accuracy respectively which outperformed the previous performances by notable margins.","PeriodicalId":255498,"journal":{"name":"2022 25th International Conference on Computer and Information Technology (ICCIT)","volume":"11 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132576407","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Images of skin lesions may be used to detect this virus, which is a reliable method for identifying the pox virus group. However, early identification and prediction are difficult due to the virus’s resemblance to other pox viruses. An intelligent computer-aided detection system may be a great alternative to relying on labor-intensive human identification. Therefore, in this research an machine learning and deep learning classification method for monkeypox prediction has been proposed and trained, and tested over 1300 skin lesion images. A comparative analysis of machine learning algorithms (K-NN and SVM) and Deep learning algorithms (Vision Transformer, RestNet50) to establish the efficacy of this study. Layered Convolutional Neural Network (CNN) with transfer learning and pretrained models such as RestNet50 integrated, together with customized hyperparameters for extracting the features from the input images. The feed-forward, which is also completely integrated, helped the algorithm divide the visuals into two categories–chickenpox and monkeypox. Among the ML model, the K-NN achieves the best accuracy of 84%. However, The Vision Transformer(ViT) outperforms the other models with an accuracy of 93%. In Addition to it, we analyze our pretrained model to achieve the desired outcome based on the relevant existing model as already established to the end user.
{"title":"Vision Transformer based Deep Learning Model for Monkeypox Detection","authors":"Dipanjali Kundu, Umme Raihan Siddiqi, Md. Mahbubur Rahman","doi":"10.1109/ICCIT57492.2022.10054797","DOIUrl":"https://doi.org/10.1109/ICCIT57492.2022.10054797","url":null,"abstract":"Images of skin lesions may be used to detect this virus, which is a reliable method for identifying the pox virus group. However, early identification and prediction are difficult due to the virus’s resemblance to other pox viruses. An intelligent computer-aided detection system may be a great alternative to relying on labor-intensive human identification. Therefore, in this research an machine learning and deep learning classification method for monkeypox prediction has been proposed and trained, and tested over 1300 skin lesion images. A comparative analysis of machine learning algorithms (K-NN and SVM) and Deep learning algorithms (Vision Transformer, RestNet50) to establish the efficacy of this study. Layered Convolutional Neural Network (CNN) with transfer learning and pretrained models such as RestNet50 integrated, together with customized hyperparameters for extracting the features from the input images. The feed-forward, which is also completely integrated, helped the algorithm divide the visuals into two categories–chickenpox and monkeypox. Among the ML model, the K-NN achieves the best accuracy of 84%. However, The Vision Transformer(ViT) outperforms the other models with an accuracy of 93%. In Addition to it, we analyze our pretrained model to achieve the desired outcome based on the relevant existing model as already established to the end user.","PeriodicalId":255498,"journal":{"name":"2022 25th International Conference on Computer and Information Technology (ICCIT)","volume":"93 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132111530","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-12-17DOI: 10.1109/ICCIT57492.2022.10055374
M. M. Hossain, Munira Akter Mou, Mst. Najmun Nahar Oishi
Recently, automated methods for disease identification have gained popularity. Many research studies use different languages for disease detection systems. We describe a disease identification method using our own developed Bengali symptoms-based disease prediction dataset that is written in the Bengali language. We have designed a disease prediction system using a transfer learning technique where we use a transformer network-based pertained model called BERT (Bidirectional Encoder Representations from Transformers). We have used the Hugging Face Transformer and then further fine-tune the model on our relatively smaller dataset. These transformer network-based deep learning techniques help us to achieve a satisfactory accuracy of 93.75%, which is good enough to identify most of the diseases using our Bengali disease dataset. The aim of the research is to use Bangla medical text data and a transfer-network based pertained model to accurately identify relevant diseases from symptoms. This will allow patients to treat their disease instantly and ensure effective disease prediction.
最近，疾病识别的自动化方法得到了普及。许多研究在疾病检测系统中使用不同的语言。我们使用我们自己开发的以孟加拉语编写的基于孟加拉症状的疾病预测数据集描述了一种疾病识别方法。我们使用迁移学习技术设计了一个疾病预测系统，其中我们使用了一个基于变压器网络的相关模型BERT(双向编码器表示从变压器)。我们使用了hug Face Transformer，然后在相对较小的数据集上进一步微调模型。这些基于变压器网络的深度学习技术帮助我们达到了令人满意的93.75%的准确率，这足以使用我们的孟加拉疾病数据集识别大多数疾病。本研究的目的是利用孟加拉医学文本数据和基于传输网络的相关模型，从症状中准确识别相关疾病。这将使患者能够立即治疗他们的疾病，并确保有效的疾病预测。
{"title":"Symptoms Based Disease Prediction from Bengali Text Using Transformer Network Based Pretrained Model","authors":"M. M. Hossain, Munira Akter Mou, Mst. Najmun Nahar Oishi","doi":"10.1109/ICCIT57492.2022.10055374","DOIUrl":"https://doi.org/10.1109/ICCIT57492.2022.10055374","url":null,"abstract":"Recently, automated methods for disease identification have gained popularity. Many research studies use different languages for disease detection systems. We describe a disease identification method using our own developed Bengali symptoms-based disease prediction dataset that is written in the Bengali language. We have designed a disease prediction system using a transfer learning technique where we use a transformer network-based pertained model called BERT (Bidirectional Encoder Representations from Transformers). We have used the Hugging Face Transformer and then further fine-tune the model on our relatively smaller dataset. These transformer network-based deep learning techniques help us to achieve a satisfactory accuracy of 93.75%, which is good enough to identify most of the diseases using our Bengali disease dataset. The aim of the research is to use Bangla medical text data and a transfer-network based pertained model to accurately identify relevant diseases from symptoms. This will allow patients to treat their disease instantly and ensure effective disease prediction.","PeriodicalId":255498,"journal":{"name":"2022 25th International Conference on Computer and Information Technology (ICCIT)","volume":"77 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134054328","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-12-17DOI: 10.1109/ICCIT57492.2022.10054937
A. Khan, Fida Kamal, Nuzhat Nower, Tasnim Ahmed, Tareque Mohmud Chowdhury
In public health surveillance, the identification of Personal Health Mentions (PHM) is an essential initial step. It involves examining a social media post that mentions an illness and determining whether the context of the post is about an actual person facing the illness or not. When attempting to determine how far a disease has spread, the monitoring of such public posts linked to healthcare is crucial, and numerous datasets have been produced to aid researchers in developing techniques to handle this. Unfortunately, social media posts tend to contain links, emojis, informal phrasing, sarcasm, etc., making them challenging to work with. To handle such issues and detect PHMs directly from social media posts, we propose a few transformer-based models and compare their performances. These models have not undergone a thorough evaluation in this domain, but are known to perform well on other language-related tasks. We trained the models on an imbalanced dataset produced by collecting a large number of public posts from Twitter. The empirical results show that we have achieved state-of-the-art performance on the dataset, with an average F1 score of 94.5% with the RoBERTa-based classifier. The code used in our experiments is publicly available1.
{"title":"An Evaluation of Transformer-Based Models in Personal Health Mention Detection","authors":"A. Khan, Fida Kamal, Nuzhat Nower, Tasnim Ahmed, Tareque Mohmud Chowdhury","doi":"10.1109/ICCIT57492.2022.10054937","DOIUrl":"https://doi.org/10.1109/ICCIT57492.2022.10054937","url":null,"abstract":"In public health surveillance, the identification of Personal Health Mentions (PHM) is an essential initial step. It involves examining a social media post that mentions an illness and determining whether the context of the post is about an actual person facing the illness or not. When attempting to determine how far a disease has spread, the monitoring of such public posts linked to healthcare is crucial, and numerous datasets have been produced to aid researchers in developing techniques to handle this. Unfortunately, social media posts tend to contain links, emojis, informal phrasing, sarcasm, etc., making them challenging to work with. To handle such issues and detect PHMs directly from social media posts, we propose a few transformer-based models and compare their performances. These models have not undergone a thorough evaluation in this domain, but are known to perform well on other language-related tasks. We trained the models on an imbalanced dataset produced by collecting a large number of public posts from Twitter. The empirical results show that we have achieved state-of-the-art performance on the dataset, with an average F1 score of 94.5% with the RoBERTa-based classifier. The code used in our experiments is publicly available1.","PeriodicalId":255498,"journal":{"name":"2022 25th International Conference on Computer and Information Technology (ICCIT)","volume":"387 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121778368","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-12-17DOI: 10.1109/ICCIT57492.2022.10055213
Asiful Islam, Sharmin Sultana Sharmee, Muhammad Nazrul Islam, Firoz Hasan, Anusha Aziz
ATMs are a type of equipment that uses a more managed method to provide customers with financial services. Almost all government and non-government banks have ATMs, allowing millions of people to get cash. For a wider adoption of ATM services in Bangladesh, ATM interfaces need to be usable, user-friendly, and easily accessible to the general public. The goal of this study is to evaluate the usability of ATM services and make design recommendations for improving their usability from human-computer interaction (HCI) perspective. To achieve these goals, the usability of five different bank ATMs currently operating in Bangladesh (DBBL, City, Brac, Islami, and AB Bank) was assessed using Heuristic Evaluation and User Evaluation approaches. The studies found that each ATM has a number of usability issues. The severity levels of these difficulties ranged from Minor usability problems (level 2) to usability catastrophe (level 4). They were primarily connected to design, help documentation, error management, user control, freedom, and the like. The survey responses showed what users really wanted based on the situation and how they interacted with the interfaces. The survey revealed that all ATM interfaces have various usability concerns to fix.
{"title":"Evaluating the Human-Computer Interaction Problems with ATM Interfaces","authors":"Asiful Islam, Sharmin Sultana Sharmee, Muhammad Nazrul Islam, Firoz Hasan, Anusha Aziz","doi":"10.1109/ICCIT57492.2022.10055213","DOIUrl":"https://doi.org/10.1109/ICCIT57492.2022.10055213","url":null,"abstract":"ATMs are a type of equipment that uses a more managed method to provide customers with financial services. Almost all government and non-government banks have ATMs, allowing millions of people to get cash. For a wider adoption of ATM services in Bangladesh, ATM interfaces need to be usable, user-friendly, and easily accessible to the general public. The goal of this study is to evaluate the usability of ATM services and make design recommendations for improving their usability from human-computer interaction (HCI) perspective. To achieve these goals, the usability of five different bank ATMs currently operating in Bangladesh (DBBL, City, Brac, Islami, and AB Bank) was assessed using Heuristic Evaluation and User Evaluation approaches. The studies found that each ATM has a number of usability issues. The severity levels of these difficulties ranged from Minor usability problems (level 2) to usability catastrophe (level 4). They were primarily connected to design, help documentation, error management, user control, freedom, and the like. The survey responses showed what users really wanted based on the situation and how they interacted with the interfaces. The survey revealed that all ATM interfaces have various usability concerns to fix.","PeriodicalId":255498,"journal":{"name":"2022 25th International Conference on Computer and Information Technology (ICCIT)","volume":"61 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130160483","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-12-17DOI: 10.1109/ICCIT57492.2022.10054952
Md. Motaleb Hossen Manik, Fabliha Haque, M. Hashem, Md. Ahsan Habib, Md. Zabirul Islam, Tanim Ahmed
Sentiment analysis has gained significant interest from multiple perspectives due to the rise of user interactions on social media and the web. It assists people in choosing the best service or product by analyzing the reviews of available options. Due to the current rise in demand, Bangla sentiment analysis has gained popularity throughout the research community. However, almost all Bangla sentiment analysis research has focused on a single approach, which has created a research gap in this domain. Therefore, this paper proposes a hybrid framework to perform sentiment analysis on Bangla texts that combines machine learning and a rule-based approach. This research starts with the machine learning approach and then integrates its intermediate result with the result of a newly proposed rule-based approach to produce the final sentiment of reviews. The experimental analysis states that the proposed hybrid framework outperforms the previous works with an accuracy of 95.54%, which assures its efficacy.
{"title":"A Hybrid Framework for Sentiment Analysis from Bangla Texts","authors":"Md. Motaleb Hossen Manik, Fabliha Haque, M. Hashem, Md. Ahsan Habib, Md. Zabirul Islam, Tanim Ahmed","doi":"10.1109/ICCIT57492.2022.10054952","DOIUrl":"https://doi.org/10.1109/ICCIT57492.2022.10054952","url":null,"abstract":"Sentiment analysis has gained significant interest from multiple perspectives due to the rise of user interactions on social media and the web. It assists people in choosing the best service or product by analyzing the reviews of available options. Due to the current rise in demand, Bangla sentiment analysis has gained popularity throughout the research community. However, almost all Bangla sentiment analysis research has focused on a single approach, which has created a research gap in this domain. Therefore, this paper proposes a hybrid framework to perform sentiment analysis on Bangla texts that combines machine learning and a rule-based approach. This research starts with the machine learning approach and then integrates its intermediate result with the result of a newly proposed rule-based approach to produce the final sentiment of reviews. The experimental analysis states that the proposed hybrid framework outperforms the previous works with an accuracy of 95.54%, which assures its efficacy.","PeriodicalId":255498,"journal":{"name":"2022 25th International Conference on Computer and Information Technology (ICCIT)","volume":"44 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129268251","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：