Aslan, Serpil; Kartoğlu, Celal Şamil; Can, Ümit

Multilingual sentiment analysis remains challenging because of cross-linguistic differences, uneven data availability, and potential shifts in meaning introduced during translation. Rather than proposing a new model, this study examines the performance of translation-based pipelines and ensemble Transformer models in multilingual settings. The primary objective is to investigate differences in downstream sentiment classification performance between translation systems within a consistent evaluation framework. Texts in Arabic, Chinese, French, and Italian are standardized through label harmonization and basic preprocessing and subsequently translated into English using Google Translate and the open-source LibreTranslate. Sentiment classification is performed using three Transformer-based models: DeBERTa-v3, RoBERTa, and BERTweet. The experiments are formulated as a binary polarity-classification task comprising positive and negative instances, with neutral samples excluded because their annotations are not directly comparable across the heterogeneous source datasets. Predictions from the individual models are combined through majority voting without introducing a new ensemble method. Experiments on the resulting binary datasets demonstrate that the ensemble achieves the highest mean weighted F1-score, with low fold-to-fold variation under both translation pipelines. Google Translate yields higher and more consistent classification performance across the evaluated subsets, whereas LibreTranslate exhibits greater variation. Given that the datasets differ in domain and genre, these variations cannot be attributed solely to linguistic differences. Overall, the findings reveal differences in downstream sentiment classification performance between the two translation pipelines and indicate that the ensemble combines high mean performance with low fold-to-fold variation within the evaluated binary setting.

İncele

Utku, Anıl; Can, Umit; Aslan, Serpil

In this article the affiliation details for Serpil Aslan were incorrectly given as ‘Software Engineering Department, Turgut Üzal University, 44210 Malatya, Turkey’ but should have been ‘Department of Software Engineering, Faculty of Engineering and Natural Sciences, Malatya Turgut Ozal University, Malatya, Turkey.’ The original article has been corrected.

AtıfScopus1·OpenAlex2

İncele

Utku, Anıl; Can, Umit; Aslan, Serpil

Today, hate speech is widespread and persistent in various forms on social networking platforms, targeting different minority groups. These attacks can be carried out using various factors such as racial, religious, gender, and physical disability, etc. Considering the number of people and their interactions, social networks are the most important channels through which these discourses spread. The social network structure is considered a set of nodes and edges and is very suitable for the graph structure. The multidimensional structure of social networks carries social network data from Euclidean space to non-Euclidean space. In non-Euclidean space, the graph structure is used to represent data effectively. In this respect, solving the hate speech problem with graph-based methods in a complex dimensional space can produce more impressive results. In this study, a powerful method based on the Graph Convolutional Network (GCN) model, which is rarely used in this field, was proposed for the detection of hateful Twitter users in social networks. Well-known machine learning methods were used to measure the performance of this method. According to the results obtained, the proposed GCN model gave the most successful result.

AtıfScopus15·WoS10·OpenAlex16

İncele

Aslan, Serpil; Kaya, B.

The traditional projection models in the bipartite networks involve many node pairs that consist of weak relationships. These node pairs lead to poor quality predictions as well as high computation time. In this paper, to cope with these problems, we firstly propose a novel projection model, which is called “Strengthened Projection Model”. Then, to predict the potential links in the future, we present a new link prediction approach based on the proposed projection model. Thanks to the proposed model, the computation time is shortened, and the high probability predictions are extracted. The majority of the previous works conducted in this area have used the classical proximity measure algorithms that only take into account the current network structure, regardless of when events occur in the network evolution. To overcome these limited methods, in this paper, we also propose a novel proximity measure algorithm that considers the bipartite network evolution. To the best of our knowledge, this is the first attempt that takes into account the time-awareness in bipartite networks. To evaluate the performance of our proposed approach, we conducted experiments on the academic information network. To construct this bipartite network, we collected data from IEEE Xplore. The experimental results show that the success of the proposed method is promising.

AtıfScopus11·WoS9·OpenAlex10

İncele

Aslan, Serpil; Kaya, Mehmet

Link prediction in complex networks is a topic of high interest for many scientists that studied different areas. Many real-world complex networks have naturally bipartite structured. In this work, we proposed a link prediction method for the complex bipartite network. Recently, while link prediction in complex bipartite networks has been extensively studied, there isn't still sufficiently suitable methods for this networks. The previous works usually have some limitations. They first map the bipartite network into the unimodal projection networks. Then, they perform the link prediction using only the properties of the projection networks. However, these projection networks involve many node pairs with weak relationships when used in networks constructed from real-world data. These weak relationships lead to low probability predictions and high computation time for the algorithm. In this paper, to overcome these limitations, we first map the bipartite network into strengthening projection network model that consists of nodes that have only strong relationships. Based on this projection network, we presented the link prediction method for the complex bipartite network. We conducted extensive experiments on complex bipartite networks. The results of the experiments demonstrate that our proposed method can effectively solve the link prediction problem in complex bipartite networks.

AtıfScopus2·OpenAlex2

İncele

Aslan, Serpil; Kaya, Buket; Varol, A.; Yazici, A.; Yazici, M.A.; Varol, C.; Aygunes, G.S.; Cotur, A.

Scientific papers are mostly performed by collaboration between research groups. Researchers work in various subjects and in several research areas. Therefore, identifying a strong research group is a quite complex task. To cope with this complex task, we propose a network-based method that predicts the best appropriate collaborations for researchers in a co-authorship network. In this paper, we have constructed a weighted co-authorship network by using a real database of scientific papers. Almost all of the methods in this area use the unweighted network model. But, weighted networks are particularly important when representing the network-based properties. The constructed network consists of nodes and links which represent respectively authors and their collaborations. In these networks, if two researchers are the common authors in the same paper, they are considered connected. Typically, the link weights represent the number of papers that two authors have made collaboratively. We also use local LCP-based algorithms to calculate the betweenness and closeness of the authors in this paper. The experimental results demonstrate the success of the proposed method.

AtıfScopus3·OpenAlex2

İncele

Aslan, Serpil; Kaya, Mehmet

Recommender systems are an important field that aim to direct users more quickly to their needs. Scientific research studies in academic field are mostly carried out by collaboration between scientists. Finding new collaborations and analyzing the quality of them are quite complicated situations. Therefore, developing a system that recommend strong collaborations for scientists can be an invaluable tool. In this study, we present a hybrid recommendation method that uses both collaborative filtering and a network-based method by examining the structure of the co-Authorship network that was created by using a real database of scientific articles in computer science. In the constructed network, the nodes represent scientists and links represent the coauthorship relationships. If two scientists in the network have been co-Author of at least one article, they are considered to be connected. Many of the present recommendation methods in this area take into account only the common neighbors of the nodes, while calculating the similarity of the node pairs in the network. To overcome these limited methods, we use the local community-based similarity indexes which also take into account the similarities between the common neighbors of the nodes. When the experimental results are examined, they prove the success of the proposed method.

AtıfScopus7·WoS3·OpenAlex6

İncele

Aslan, Serpil; Kaya, Buket; Kaya, Mehmet

The link prediction problem can be used for predicting the link changes that are difficult to understand and costly in network science. So far, many different models have been proposed to predict links in the network. Most of the proposed models in the link prediction have focused on the unimodal networks. In this paper, we present a link prediction model that attempts to predict links in the large-scale bipartite networks. The proposed method consists of two stages: The first stage is the extraction of the potential link set. A limitation of the previous works that addressed the link prediction problem in bipartite networks is that only the structural properties of the projected networks that were transformed from the bipartite networks are used to extract the potential links. In this paper, in addition to the structural properties of the projected networks, we take into account the structural properties of the bipartite network. Also, we employ the strengthened projection network model instead of the classical projection network models. The second stage is the calculation of prediction score of each potential link. The classical proximity measures have several disadvantages in dynamic complex networks. When network evolution that considers valuable information between node pairs are ignored, the potential information source is not efficiently used in the link prediction. Therefore, in this stage, we propose a time-aware proximity measure based on social network evolution, instead of classical proximity measures using the current state of the network. To the best of our knowledge, this is the first effort that considers the evolution in bipartite networks. To test the performance of the proposed method, we made some experiments on two real networks that have different frame size. Experimental results show that the success of our method is promising.

AtıfScopus14·WoS5·OpenAlex9

İncele

Aslan, Serpil; Kaya, Mehmet

Recently, link prediction studies on large-scale and complex networks have particularly become the focus of interest for researchers in various scientific fields. Many complex networks created from the real-world data contain bipartite structure by nature. Bipartite networks are a kind of complex networks that represent the interactions between the different node groups. Almost all of the previous studies on link prediction in bipartite networks focus on using the properties of projection networks to predict the relations between the node pairs. In this study, a novel similarity-based link prediction method based on strengthening weighted projection is proposed to predict the potential links between the authors and the topics in the large-scale bipartite academic information network created from the real-world data. Since information loss occurs when bipartite networks are converted to unimodal networks, it should be noted that when making a link prediction in this paper, both the bipartite network and the information on strengthening unimodal network obtained from the bipartite network are used. To evaluate the proposed method, a bipartite network was first created from a real dataset consisting of authors and their work. Then the method was tested on this network. Experimental results demonstrate that it is possible to obtain faster and more accurate link prediction results by the proposed method.

AtıfScopus35·WoS25·OpenAlex34

İncele

Aslan, Serpil; Kaya, Mehmet; Adali, E

Nowadays, social networks, whose application areas are increasing day by day; Such as data mining, protein-gene interaction networks, online social networks, transport networks, academic information networks, and telecommunications networks. Many real-world complex networks have a naturel bipartite structure. In bipartite networks, which have become the focus of many researchers who have been dealing with social network analysis in recent years, connection estimation methods are the basic methods used to understand the structure of these networks. This study is a survey work that describes the connection in a bipartite network that is suitable for the nature of today's huge data networks. In this study, a bipartite network interface is an introductory guide for researchers who plan to work on this subject by examining the internal links and bipartite network projection methods.

AtıfWoS4

İncele