Scopus İndeksli Yayınlar Koleksiyonu
Permanent URI for this collectionhttps://hdl.handle.net/20.500.12573/395
Browse
20 results
Search Results
Conference Object Citation - Scopus: 1Words Speak Louder Than Actions: Decoding Emotions Through NLP(Institute of Electrical and Electronics Engineers Inc., 2024-10-26) Paksoy, Melda; Bakal, GokhanEmotion detection in text remains a significant challenge in Natural Language Processing due to human emotions' complexity and subtle nuances. This paper presents multiple experimental models for emotion classification using an up-to-date dataset curated to address 13 emotions implied in Twitter posts. We evaluated various machine learning (ML) models, including Logistic Regression, Random Forest, SVM, and XGBoost, alongside deep learning (DL) architectures such as LSTM and CNN. Our results demonstrate the efficacy of deep learning models, particularly the CNN model by achieving an impressive F1 score of 0.99. This study contributes to emotion detection capabilities, paving the way for more nuanced and accurate sentiment analysis (SA) in various text analysis applications. © 2025 Elsevier B.V., All rights reserved.Conference Object Text Classification Experiments on Contextual Graphs Built by N-Gram Series(Springer International Publishing AG, 2025) Sen, Tarik Uveys; Yakit, Mehmet Can; Gumus, Mehmet Semih; Abar, Orhan; Bakal, GokhanTraditional n-gram textual features, commonly employed in conventional machine learning models, offer lower performance rates on high-volume datasets compared to modern deep learning algorithms, which have been intensively studied for the past decade. The main reason for this performance disparity is that deep learning approaches handle textual data through the word vector space representation by catching the contextually hidden information in a better way. Nonetheless, the potential of the n-gram feature set to reflect the context is open to further investigation. In this sense, creating graphs using discriminative ngram series with high classification power has never been fully exploited by researchers. Hence, the main goal of this study is to contribute to the classification power by including the long-range neighborhood relationships for each word in the word embedding representations. To achieve this goal, we transformed the textual data by employing n-gram series into a graph structure and then trained a graph convolution network model. Consequently, we obtained contextually enriched word embeddings and observed F1-score performance improvements from 0.78 to 0.80 when we integrated those convolution-based word embeddings into an LSTM model. This research contributes to improving classification capabilities by leveraging graph structures derived from discriminative n-gram series.Conference Object Citation - Scopus: 1Semantic-Forward Relaying for 6G: Performance Boosts With ResNet-18 and GoogleNet Plus(Institute of Electrical and Electronics Engineers Inc., 2024-11-28) Erkantarci, Betul; Çoban, Mert Korkut; Bozoǧlu, Abdulkadir; Köse, AbdulkadirThis paper investigates the integration of advanced deep learning architectures, namely ResNet-18, GoogleNet and enhanced GoogleNet (GoogleNet Plus), into the Semantic-Forward (SF) relaying framework for cooperative communications in 6G networks. The SF relaying framework enhances transmission efficiency and robustness by leveraging semantic information at relay nodes. We analyze and compare the performance of these deep learning models in terms of validation accuracy, semantic accuracy, and Euclidean distance (ED) metrics on the CIFAR-10 dataset. Results indicate that ResNet-18 achieves the highest performance due to its residual learning architecture. GoogleNet Plus, incorporating Automatic Mixed Precision (AMP) training and the Adam optimizer, demonstrates improved stability and efficiency compared to the original GoogleNet. The results highlights the potential of deep learning models to enhance semantic processing capabilities in SF relaying, contributing to the development of more efficient, resilient, and adaptive cooperative communication systems in 6G networks. © 2025 Elsevier B.V., All rights reserved.Conference Object Citation - Scopus: 2Re-Exploring the Kayseri Culture Route by Using Deep Learning for Cultural Heritage Image Classification Cultural Heritage Image Classification by Using Deep Learning: Kayseri Culture Route(Association for Computing Machinery, 2024-05-25) Kevseroğlu, Ozlem; Kurban, RifatThe categorization of images captured during the documentation of architectural structures is a crucial aspect of preserving cultural heritage in digital form. Dealing with a large volume of images makes this categorization process laborious and time-consuming, often leading to errors. Introducing automatic techniques to aid in sorting would streamline this process, enhancing the efficiency of digital documentation. Proper classification of these images facilitates improved organization and more effective searches using specific terms, thereby aiding in the analysis and interpretation of the heritage asset. This study primarily focuses on applying deep learning techniques, specifically SqueezeNet convolutional neural networks (CNNs), for classifying images of architectural heritage. The effectiveness of training these networks from scratch versus fine-tuning pre-existing models is examined. In this study, we concentrate on identifying significant elements within images of buildings with architectural heritage significance of Kayseri Culture Route. Since no suitable datasets for network training were found, a new dataset was created. Transfer learning enables the use of pre-trained convolutional neural networks to specific image classification tasks. In the experiments, 99.8% of classification accuracy have been achieved by using SqueezeNet, suggesting that the implementation of the technique can substantially enhance the digital documentation of architectural heritage. © 2024 Elsevier B.V., All rights reserved.Book Part Citation - Scopus: 3ROSE: A Novel Approach for Protein Secondary Structure Prediction(Springer Science and Business Media Deutschland GmbH, 2021) Görmez, Yasin; Aydin, ZaferThree-dimensional structure of protein gives important information about protein’s function. Since it is time-consuming and costly to find the structure of protein by experimental methods, estimation of three-dimensional structures of proteins through computational methods has been an efficient alternative. One of the most important steps for the 3-D protein structure prediction is protein secondary structure prediction. Proteins which contain different number and sequences of amino acids may have similar structures. Thus, extracting appropriate input features has crucial importance for secondary structure prediction. In this study, a novel model, ROSE, is proposed for secondary structure prediction that obtains probability distributions as a feature vector by using two position specific scoring matrices obtained by PSIBLAST and HHblits. ROSE is a two-stage hybrid classifier that uses a one-dimensional bi-directional recurrent neural network at the first stage and a support vector machine at the second stage. It is also combined with DSPRED method, which employs dynamic Bayesian networks and a support vector machine. ROSE obtained comparable results to DSPRED in cross-validation experiments performed on a difficult benchmark and can be used as an alternative to protein secondary structure prediction. © 2021 Elsevier B.V., All rights reserved.Conference Object Citation - Scopus: 1PCB Component Recognition With Semi-Supervised Image Clustering(IEEE, 2021-06-09) Unal, Ahmet Emin; Tasdemir, Kasim; Bahcebasi, AkifClassification of surface mounted devices plays an important role on automated inspection systems of printed component board production. Limited number of publicly available datasets which the components are labeled and high intraclass variance in these datasets causes the supervised approches to be inefficient. In this study a deep learning method, enhanced with an unsupervised clustering system, which uses a small set of labeled data is proposed. The method compared with the current studies and the supervised systems. Most optimized setting reached high accuracy results by outrunning current classification methods.Conference Object Citation - Scopus: 8On Comparative Classification of Relevant COVID-19 Tweets(Institute of Electrical and Electronics Engineers Inc., 2021-09-15) Bakal, Gokhan; Abar, OrhanDue to the impressive information dissemination power of social networks such as Twitter, people tend to check social networks and Web pages more than other traditional news sources, including newspapers, TV news programs, or radio channels. In that sense, the information carried by the content of the shared social media posts becomes much more considerable. However, most of the posts are commonly either irrelevant or inaccurate. Besides, the more critical case than the correctness of the information is the diffusion speed on Twitter through the reply or retweet actions. These activities make the initial situation even more complicated than itself due to the unregulated nature of the social networks and the lack of an immediate verification mechanism for the correctness of the posts. When we consider the current Covid-19 pandemic period (causing the coronavirus disease), one of the most utilized information resources is Twitter except the official health administration institutions. Thereupon, examining the correctness of the information related to the Covid-19 pandemic by computational techniques (e.g., Data Mining, Machine Learning, and Deep Learning) has been gaining popularity and remains a substantial task. Hence, we mainly focused on analyzing the correctness of the posts related to the current pandemic shared on the Twitter platform. Therefore, the overall goal of this work is to classify the relevant tweets using linear and non-linear machine learning models. We achieved the best F1 performance score (99%) with the neural network model using the unigram features & threshold value of 50 among all model configurations. © 2022 Elsevier B.V., All rights reserved.Conference Object Multi-Method Text Summarization: Evaluating Extractive and BART-Based Approaches on CNN/Daily Mail(Institute of Electrical and Electronics Engineers Inc., 2025-06-27) Inal, Yasin; Bakal, Gokhan; Esit, MuhammedWith the exponential growth of digital content, efficient text summarization has become increasingly crucial for managing information overload. This paper presents a comprehensive approach to text summarization using both extractive and abstractive methods, implemented on the CNN/Daily Mail dataset. We leverage pre-trained BART (Bidirectional and AutoRegressive Transformers) models and fine-tuning techniques to generate high-quality summaries. Our approach demonstrates significant improvements, with our best model trained on 287 k samples achieving ROUGE-1 F1 scores of 0.4174, ROUGE-2 F1 scores of 0.1932, and ROUGE-L F1 scores of 0.2910. We provide detailed comparisons between extractive methods and various BART model configurations, analyzing the impact of training dataset size and model architecture on summarization quality. Additionally, we share our implementation through an opensource NLP toolkit to facilitate further research and practical applications in the field. © 2025 Elsevier B.V., All rights reserved.Conference Object Linear Vs. Non-Linear Embedding Methods in Recommendation Systems(Institute of Electrical and Electronics Engineers Inc., 2022-09-07) Gurler, Kerem; Cos¸kun, Mustafa; Karagenc, Safak; Orun, Gokhan; Kuleli Pak, Burcu Kuleli; Güngör, Vehbi Çağrı; Coskun, Mustafa; Pak, Burcu KuleliPredicting customer interest in items is very crucial in direct marketing as it can potentially boost sales. Data mining techniques are developed to predict which items a particular user might be interested in based on their purchase history or explicit feedback in form of ratings or comments. Recently, non-linear and linear methods have been developed for this purpose. In this study, we applied Neighborhood based Collaborative Filtering (CF), Matrix Factorization (MF), Singular Value Decomposition (SVD), Neural Graph CF (NGCF) and Light Graph Convolutional Network (LightGCN) on explicit user product rating data which is acquired from the online gaming and mobile entertainment platform called HADI. We compared the results of node embedding methods in terms of Precision@k, Recall@k and NDCG@k values. SVD and LightGCN showed the best test performance and SVD was significantly superior to LightGCN in terms of training speed. To further increase predictive performance of SVD, we have applied classification with Logistic Regression and Deep Random Forest on user and item embeddings created by the SVD. © 2022 Elsevier B.V., All rights reserved.Conference Object Citation - Scopus: 7Generating Emergency Evacuation Route Directions Based on Crowd Simulations With Reinforcement Learning(Institute of Electrical and Electronics Engineers Inc., 2022-09-07) Unal, Ahmet Emin; Gezer, Cengiz; Kuleli Pak, Burcu Kuleli; Güngör, Vehbi Çağrı; Pak, Burcu KuleliIn an emergency, it is vital to evacuate individuals from the dangerous environments. Emergency evacuation plan-ning ensures that the evacuation is safe and optimal in terms of evacuation time for all of the people in evacuation. To this end, the computer-enabled evacuation simulation systems are used to generate optimal routes for the evacuees. In this paper, a dynamic emergency evacuation route generator has been proposed based on indoor plans of the building and the locations of the evacuees. To generate the optimal routes in real-time, a reinforcement learning algorithm (proximal policy optimization) is presented. Comparative performance results show that the proposed model is successful for evacuating the individuals from the building in different scenarios. © 2022 Elsevier B.V., All rights reserved.