ABSTRACT This editorial is for the Special Issue on the 8th High‐Performance Computing Conference (BAŞARIM 2024), held on May 15–17, 2024, at the Middle East Technical University Culture and Convention Center in Ankara.

In this work, we tackle the multi-modal irony detection problem and propose a lightweight approach that combines pretrained multi-modal and text-based smaller language models. The proposed approach makes use of generated image captions and original text content to fine-tune a transformer based model for text classification. In the proposed solution, Vision-Language Model (VLM) is used for generating image captions, without any fine-tuning. On two benchmark datasets, the experiments demonstrate the benefit of the proposed method against the baseline solutions.

This study evaluates the effectiveness of Vision Language Models (VLMs) in representing and utilizing multimodal content for fact-checking. To be more specific, we investigate whether incorporating multimodal content improves performance compared to text-only models and how well VLMs utilize text and image information to enhance misinformation detection. Furthermore we propose a probing classifier based solution using VLMs. Our approach extracts embeddings from the last hidden layer of selected VLMs and inputs them into a neural probing classifier for multi-class veracity classification. Through a series of experiments on two fact-checking datasets, we demonstrate that while multimodality can enhance performance, fusing separate embeddings from text and image encoders yielded superior results compared to using VLM embeddings. Furthermore, the proposed neural classifier significantly outperformed KNN and SVM baselines in leveraging extracted embeddings, highlighting its effectiveness for multimodal fact-checking.

The rapid dissemination of misinformation through social media increased the importance of automated fact-checking. Furthermore, studies on what deep neural models pay attention to when making predictions have increased in recent years. While significant progress has been made in this field, it has not yet reached a level of reasoning comparable to human reasoning. To address these gaps, we propose a multi-task explainable neural model for misinformation detection. More specifically, this work formulates an explanation generation process of the model’s veracity prediction as a text summarization problem. Additionally, the performance of the proposed model is discussed on publicly available datasets and the findings are evaluated with related studies.

The rapid spread of misinformation through social media platforms has raised concerns regarding its impact on public opinion. While misinformation is prevalent in other languages, the majority of research in this field has concentrated on the English language. Hence, there is a scarcity of datasets for other languages, including Turkish. To address this concern, we have introduced the FCTR dataset, consisting of 3238 real-world claims. This dataset spans multiple domains and incorporates evidence collected from three Turkish fact-checking organizations. Additionally, we aim to assess the effectiveness of cross-lingual transfer learning for low-resource languages, with a particular focus on Turkish. We demonstrate in-context learning (zero-shot and few-shot) performance of large language models in this context. The experimental results indicate that the dataset has the potential to advance research in the Turkish language.

In the big data era, automated index selection and recommendation has been an important research problem to improve the data access efficiency. Previous efforts on artificial intelligence based database index selection have focused on relational databases. In this work, we consider the automated index selection for NoSQL databases and investigate the feasibility of supervised learning and reinforcement learning based solutions. The experiments conducted on the YCSB dataset show that reinforcement learning improves index selection performance as in relational databases, and supervised learning gives promising results and can be considered applicable under sufficient amount of training data.

Smartphones play a crucial role in daily activities, however, situationally-induced impairments and disabilities (SIIDs) can easily be experienced depending on the context. Previous studies explored the effect of context but mainly done in controlled environments with limited research done in the wild. In this article, we present an in-situ remote user study with 48 participants’ keyboard interaction on smartphones including the performance and context details. We first propose an automated approach for error detection by combining approaches introduced in the literature and with a follow-up study, show that the accuracy of error detection is improved. We then investigate the effect of context on the typing performance based on five dimensions: environment, mobility, social, multitasking, and distraction, and reveal that the context affects participants’ error rate significantly but with individual differences. Our main contribution is providing empirical evidence with an in-situ study showing the effect of context on error rate.

In social networks, trust is a fundamental notion affecting the nature and the strength of ties between individuals. It is also a piece of useful auxiliary information for improving the performance of recommendation systems. The number of ratings given by a user is minimal compared to all items in popular, widely-used e-commerce sites. Therefore, the user-item matrix that is used in collaborative filtering suffers from data sparsity, resulting in poor recommendation quality. Another issue is the cold start problem, which occurs for the inclusion of new users and new items to the system. Trust notion is helpful for alleviating the effect of these problems by providing additional relationships between the users and pointing out strong relationships. Information as to the trust between users can be explicitly available. However, such information is not widely available, and hence implicit trust models have been employed. This work analyzes two sub-problems under trust modeling for recommendation: (1) What is the relationship between explicit and implicit trust scores, are they replaceable? (2) Can we model explicit trust in a trust network? For the first problem, we present an implicit trust model and analyze the compatibility of implicit and explicit trust scores. For the second problem, we model explicit trust modeling as a link prediction problem and analyze the performance of the prediction models we generate on a set of benchmark data sets.

Abstract Learning to rank is a supervised learning problem that aims to construct a ranking model for the given data. The most common application of learning to rank is to rank a set of documents against a query. In this work, we focus on point‐wise learning to rank , where the model learns the ranking values. Multivariate adaptive regression splines (MARS) and conic multivariate adaptive regression splines (CMARS) are supervised learning techniques that have been proven to provide successful results on various prediction problems. In this article, we investigate the effectiveness of MARS and CMARS for point‐wise learning to rank problem. The prediction performance is analyzed in comparison to three well‐known supervised learning methods, artificial neural network (ANN), support vector machine, and random forest for two datasets under a variety of metrics including accuracy, stability, and robustness. The experimental results show that MARS and ANN are effective methods for learning to rank problem and provide promising results.

Named Entity Recognition (NER) is an information extraction task that aims to automatically identify named entities in a given text. Named entities are special types of nouns or noun groups that refer to specific entities including as person, location, organization, date, time, money and percentage. NER also facilitates various other Natural Language Processing (NLP) related tasks such as summarization and question answering. It is a vastly studied problem especially on English texts, however number of NER studies on Turkish is very limited. Being a morphologically rich language, Turkish has an agglutinative structure and hence automated analysis and information extraction performance is generally lower than those on English. The previous studies mostly use conventional supervised learning and sequence tagging methods, such as Conditional Random Fields (CRF). Only few studies use deep neural models for NER problem on Turkish texts. In this work, we particularly focus on the recent neural model, Google BERT, and analyze its performance on Turkish texts. In addition to fully trained BERT model, we investigate the performance of different training levels from fully trained to fully pre-trained.

The need for a variety of auxiliary analytical tools to enhance marine safety and marine status awareness has been expressed by various platforms. The information that has been published while cruising is a rich resource for movement analysis of ships. Automatic Identification System (AIS), which is widely used in vessels, broadcasts information including the type of ship, identity number, state, destination, estimated time of arrival (ETA), location, speed, direction, and cargo. In this paper, to aid maritime operators, we work on arrival port, arrival time, and next position prediction on AIS messages, and propose three different approaches for the prediction of marine vessel movement. The experiments conducted against conventional supervised learning approaches reveal the improvement of the proposed solutions.

There are over 1900 cryptocurrencies trading in the market as of September 2020 and the number is rapidly growing. In the current crypto scene, cryptocurrencies are seen as investment vehicles by many, yet every crypto asset is designed to operate in a specific sector within a pre-defined business model. In addition to sector, there are various characteristics and factors that differentiate one crypto asset from another. Crypto investors can leverage these factors and characteristics and use these indicators to create different trading strategies. In our work, in order to guide the decision-making process for investors and to help them analyse crypto assets in a holistic manner, we classify the crypto assets under various characteristics, aiming towards a crypto asset taxonomy. In this paper, we focus on automated annotation of the cryptocurrencies in terms of sector, transaction anonymity and asset type through the public information. Though the information we utilise is public, it is scattered around quite a vast number of sources in different formats. Therefore, we generated an annotated dataset by collecting information from various sources. We utilised several supervised learning algorithms, including both traditional ones and more recent neural models, and analyzed the classification performance for the three aspects.

In this paper, we present a framework for predicting the personality traits by analyzing tweets written in Turkish. The prediction model is constructed with a clustering based approach. Since the model is based on linguistic features, it is language specific. The prediction model uses features applicable to Turkish language and related to writing style of Turkish Twitter users. Our approach uses anonymous BIG5 questionnaire scores of volunteer participants as the ground truth in order to generate personality model from Twitter posts. Experiment results show that constructed model can predict personality traits of Turkish Twitter users with relatively small errors.

Automated Telling Machines (ATM) are one of the prominent services of the banks, which facilitate daily banking operations. Among the offered services, money withdrawal is a very basic functionality of ATMs. For the banks, it is important to manage the amount of money to be loaded in ATMs. Hence, prediction of the withdrawal amount is an important step in the ATM management. In this work, we investigate the performance of deep learning techniques for ATM money withdrawal amount prediction problem. The problem is defined in two ways: predicting the amount of money to be withdrawn, and predicting the class label of the withdrawal amount. Our data set includes daily total withdrawal amounts together with the date information. In our experiments, we further analyzed the effect of additional information, such as size of history window, weather condition, currency rate and location. For numeric value prediction task, we compared the prediction performance with statistical models of ARIMA and SARIMA. The experiments show that the neural architectures are feasible for the task of withdrawal amount prediction. They can especially provide high accuracy results when the problem is modeled as a class label prediction task.

Event detection is a popular research problem, aiming to detect events from online data sources with least possible delay. Most of the previous work focus on analyzing textual content such as social media postings to detect happenings. In this work, we consider event detection as a change detection problem in network structure, and propose a method that detects change in community structure extracted from communication network. We study three versions of the method based on different change models. Experimental analysis on benchmark data set reveals that change in the community can be used as an indication of an event.

A recent “third wave” of neural network (NN) approaches now delivers state-of-the-art performance in many machine learning tasks, spanning speech recognition, computer vision, and natural language processing. Because these modern NNs often comprise multiple interconnected layers, work in this area is often referred to as deep learning . Recent years have witnessed an explosive growth of research into NN-based approaches to information retrieval (IR). A significant body of work has now been created. In this paper, we survey the current landscape of Neural IR research, paying special attention to the use of learned distributed representations of textual units. We highlight the successes of neural IR thus far, catalog obstacles to its wider adoption, and suggest potentially promising directions for future research.

Learning to rank is a supervised learning problem that aims to construct a ranking model. The most common application of learning to rank is to rank a set of documents against a query. In this work, we focus on pointwise approach and compare the performances of four computational methods in developing ranking models using several criteria such as accuracy, stability and robustness. The experimental results show that Multivariate Adaptive Regression Splines (MARS) and Artificial Neural Networks (ANN) are effective methods for learning to rank problem and provide promising results.

The web provides a suitable media for users to post comments on different topics. In most of such content, authors express different opinions on different features or aspects of the topic. In aspect based sentiment analysis, it is analyzed as to for which aspect which opinion is expressed. Once aspects are available, the next important step is to match aspects with correct sentiments. In this work, we investigate enhancements for two cases in matching step: extracting implicit aspects, and sentiment words whose polarity depends on the aspect. The techniques are applied on Turkish informal texts collected from a products forum. Experimental evaluation shows that additional steps applied for these cases improve the accuracy of aspect based sentiment analysis.

This paper presents efficient data structures and a pruning technique in order to improve the efficiency of high utility sequential pattern mining. CRoM (Cumulated Rest of Match) based upper bound, which is a tight upper bound on the utility of the candidates is proposed in order to perform more conservative pruning before candidate pattern generation in comparison to the existing techniques. In addition, an efficient algorithm, HuspExt (High Utility Sequential Pattern Extraction), is presented which calculates the utilities of the child patterns based on that of the parents'. Substantial experiments on both synthetic and real datasets from different domains show that, the solution efficiently discovers high utility sequential patterns under low thresholds.

In recent years, using cell phone log data to model human mobility patterns became an active research area. This problem is a challenging data mining problem due to huge size and non-uniformity of the log data, which introduces several granularity levels for the specification of temporal and spatial dimensions. This paper focuses on the prediction of the location of the next activity of the mobile phone users. There are several versions of this problem. In this work, we have concentrated on the following three problems: predicting the location and the time of the next user activity, predicting the location of the next activity of the user when the location of the user changes, and predicting both the location and the time of the activity of the user when the user's location changes. We have developed sequential pattern mining-based techniques for these three problems and validated the success of these methods with real data obtained from one of the largest mobile phone operators in Turkey. Our results are very encouraging, since we were able to obtain quite high accuracy results under small prediction sets.

The pervasiveness of location-acquisition technologies enable location-based social networks (LBSN) to become increasingly popular in recent years. Users are able to check-in their current location and share information with other users through these networks. LBSN check-in data can be used for the benefit of users by providing personalized recommendations. There are several location recommendation algorithms that employ LBSN data in the literature. However, there are few number of proposed activity recommendation algorithms. In this paper, we propose a random walk based context-aware activity recommendation algorithm, namely RWCAR, for LBSNs. RWCAR considers the current context (i.e. social relations, personal preferences, and current location) of the user to provide recommendations. We propose a graph model for representing LBSN data that contains users, locations and activities. We build a graph according to the current context of the user depending on this LBSN model. A random walk approach is employed to predict the recommendation scores of the activities. A list of activities are recommended in decreasing order of calculated recommendation score. In experimental evaluation, we compare RWCAR with friend-based, expert-based and popularity-based activity recommendation algorithms. The proposed algorithm performs better in terms of activity recommendation accuracy in all of the experiments.

Integration of the wind power into the existing transmission grid is an important issue due to discontinuous and volatile behavior of wind. Moreover, the power plant owners need reliable information about day-ahead power production for market operations. Therefore, wind power forecasting approaches have been gaining importance in renewable energy research area. The Wind Power Monitoring and Forecast System for Turkey (RITM) currently monitors a growing number of wind power plants in Turkey, and uses wind power measurements in addition to different numerical weather predictions to generate short-term power forecasts. Forecasting models of RITM give considerably good results individually. However, forecast combination approaches are frequently used in order not to rely on a single forecast model, and also utilize forecast diversification. In this paper, an analysis of wind power domain and the current wind power forecasting methods of RITM are presented. Then, three main forecast combination approaches, namely Lp-norm based combination, FSS (Fuzzy Soft Sets) based combination and tree-based combination, are proposed to provide better forecasts. These combination methods have been verified on forecasts data of RITM in terms of normalized mean absolute error (NMAE) metric. The experimental results show that all of the applied combination methods give lower NMAE rates for most of the wind power plants compared to individual forecasts.

As the result of increasing population and growing technological activities, nonrenewable energy sources, which are the main energy providers, are diminishing day by day. Due to this factor, efforts on efficient utilization of renewable energy sources have increased all over the world. Wind is one of the most significant alternative energy resources. However, in comparison with other renewable energy sources, it is so variable that there is a need for estimating and planning of wind power generation. In this paper, a new statistical short-term (up to 48 h) wind power forecast model, namely statistical hybrid wind power forecast technique (SHWIP), is presented. In the proposed model, weather events are clustered with respect to the most important weather forecast parameters. It also combines the power forecasts obtained from three different numerical weather prediction (NWP) sources and produces a hybridized final forecast. The proposed model has been in operation at the Wind Power Monitoring and Forecast System for Turkey (RITM), and the results of the new model are compared with well-known statistical models and physical models in the literature. The most important characteristics of the proposed model is the need for a lesser amount of historical data while constructing the mathematical model compared with the other statistical models such as artificial neural networks (ANN) and support vector machine (SVM). To produce a reliable forecast, ANN and SVM need at least 1 year of historical data; on the other hand, the proposed SHWIP method's results are applicable even under 1 month of training data, and this is an important feature for the forecast of the newly established wind power plants (WPPs).

High utility sequential pattern mining has been considered as an important research problem and a number of relevant algorithms have been proposed for this topic. The main challenge of high utility sequential pattern mining is that, the search space is large and the efficiency of the solutions is directly affected by the degree at which they can eliminate the candidate patterns. Therefore, the efficiency of any high utility sequential pattern mining solution depends on its ability to reduce this big search space, and as a result, lower the computational complexity of calculating the utilities of the candidate patterns. In this paper, we propose efficient data structures and pruning technique which is based on Cumulated Rest of Match (CRoM) based upper bound. CRoM, by defining a tighter upper bound on the utility of the candidates, allows more conservative pruning before candidate pattern generation in comparison to the existing techniques. In addition, we have developed an efficient algorithm, High Utility Sequential Pattern Extraction (HuspExt), which calculates the utilities of the child patterns based on that of the parents'. Substantial experiments on both synthetic and real datasets from different domains show that, the proposed solution efficiently discovers high utility sequential patterns from large scale datasets with different data characteristics, under low utility thresholds.

Use of web services is one of the most rapidly developing technologies. Since web services are defined by XML- based standards to overcome platform dependency, they are very eligible to integrate with each other in order to establish new services. This composition enables us to reuse existing services, which results in less cost and time consumption. One of the recent problems with web service composition is to maximize the overall Quality of Service (QoS) of the composed service. Most common elements of QoS are response time, availability, reliability, throughput and cost (price). Since the selection of the optimal execution plan that maximizes the composition's overall QoS is a NP hard problem, applying optimization techniques is very popular. In this work, we propose an improved Genetic Algorithm based approach to optimize the overall QoS of the composed service. Experimental results indicate improvement for QoS of the composition built by the proposed methods.

Although Inductive Logic Programming (ILP)-based concept discovery systems have applications in a wide range of domains, they still suffer from scalability and efficiency issues. One of the reasons for the efficiency problem is the high number of query executions necessary in the concept discovery process. Owing to the refinement operator of ILP-based concept discovery systems, these queries repeat frequently. In this work, we propose a method to improve the look-up table hit ratio for repeating queries of ILP-based concept discovery systems with memoization capabilities. The proposed method introduces modifications on search space evaluation and the covering steps of such systems so that query results of the previous iterations can be exploited. Experimental results show that the proposed method improves the hash table hit ratio of ILP-based concept discovery systems with an affordable cost of extra memory consumption.

Sentiments and opinions expressed in Web pages towards objects, entities, and products constitute an important portion of the textual content available in the Web. In the last decade, the analysis of such content has gained importance due to its high potential for monetization. Despite the vast interest in sentiment analysis, somewhat surprisingly, the discovery of sentimental or opinionated Web content is mostly ignored. This work aims to fill this gap and addresses the problem of quickly discovering and fetching the sentimental content present in the Web. To this end, we design a sentiment-focused Web crawling framework. In particular, we propose different sentiment-focused Web crawling strategies that prioritize discovered URLs based on their predicted sentiment scores. Through simulations, these strategies are shown to achieve considerable performance improvement over general-purpose Web crawling strategies in discovery of sentimental Web content.

This paper aims to enhance event detection methods in a micro-blogging platform, namely Twitter. The enhancement technique we propose is based on lexico-semantic expansion of tweet contents while applying document similarity and clustering algorithms. Considering the length limitations and idiosyncratic spelling in Twitter environment, it is possible to take advantage of word similarities and to enrich texts with similar words. The semantic expansion technique we implement is based on syntagmatic and paradigmatic relationships between words, extracted from their co-occurrence statistics. As our technique does not depend on an existing ontology or a lexicon database such as Word Net, it should be applicable for any language. The proposed technique is applied on a tweet set collected for three days from the users in Turkey. The results indicate earlier detection of events and improvements in accuracy.

The sentiments and opinions that are expressed in web pages towards objects, entities, and products constitute an important portion of the textual content available in the Web. Despite the vast interest in sentiment analysis and opinion mining, somewhat surprisingly, the discovery of the sentimental or opinionated web content is mostly ignored. This work aims to fill this gap and address the problem of quickly discovering and fetching the sentimental content present in the Web. To this end, we design a sentiment-focused web crawling framework for faster discovery and retrieval of such content. In particular, we propose different sentiment-focused web crawling strategies that prioritize discovered URLs based on their predicted sentiment scores. Through simulations, these strategies are shown to achieve considerable performance improvement over general-purpose web crawling strategies in discovering sentimental content.

In this paper, we propose a new model to integrate additional data, which is obtained from geospatial resources other than original data set in order to improve Location/Activity recommendations. The data set that is used in this work is a GPS trajectory of some users, which is gathered over 2 years. In order to have more accurate predictions and recommendations, we present a model that injects additional information to the main data set and we aim to apply a mathematical method on the merged data. On the merged data set, singular value decomposition technique is applied to extract latent relations. Several tests have been conducted, and the results of our proposed method are compared with a similar work for the same data set.

In this work, we propose a new QoS evaluation method for web services. The proposed method aims to keep track of web services in terms of QoS parameters and to generate QoS values automatically when necessary. It is realized as a part of a Web service discovery system that is a collection of domain specific service discovery nodes.

By the increase in number of available Web services over the internet, accessing a required web service becomes a challenging task. One solution to the problem is using distributed web service discovery system that is composed of domain specific sub service discoverers. Using ontology is the most common practice to specify domain knowledge. Domain specific subsystems form an unstructured P2P network that connects nodes to each other in order to facilitate cross domain web service query routings. In this paper, we present a networking infrastructure of a distributed web service search system and its ontology-based routing method, We evaluate the performance of ontology-based routing method by using both complete and brief version of ontology of the domain specific discoverer under basic and normalized semantic similarity measures in terms of traffic improvement, and accuracy of the forwarded web services. Experimental evaluation shows that the traffic is reduced by 50% with the web service forwarding precision of 90%.

As the number of available web services published in registries and on web sites increases, web service discovery becomes a challenging task. One solution to the problem is to use distributed web service search system composed of domain specific sub service discoverers. Using ontology is the most common practice to specify domain knowledge. However, an important problem at this point is the lack of semantic annotation for currently available web services. For this reason, there is a strong need for a mechanism for semantic annotation of unannotated services. In this work, we propose a web service semantic annotation method that uses lexicon-based alignment. Lexicon-based alignment considers the different senses of the words, and hence it can find the association between the service and the ontology more accurately than previous alignment techniques.

Process mining is a branch of data mining that aims to discover process model from the event logs. In this study, we propose a hybrid approach to process mining in such a way that, "from-to chart" is used as the frontend to monitor the transitions among activities of a realistic event log. Another novelty of this study is developed evaluation metrics, which are used for finding immediate successors in order to convert these raw relations into dependency/frequency graph.