filter by: Publication Year
(Descending) Articles
Information Sciences (00200255) 692
Open relation extraction is a critical task in natural language processing aimed at automatically extracting relations between entities in open-domain corpora. Most existing systems focus on extracting binary relations (relations between two entities) while extracting more complex n-ary relations (involving more than two entities) remains a significant challenge. Additionally, many previous systems rely on hand-crafted patterns and natural language processing tools, which result in error accumulation and reduced accuracy. The current study proposes a novel approach to open n-ary relation extraction that leverages recent advancements in deep learning architectures. This approach addresses the limitations of existing open relation extraction systems, particularly their reliance on hand-crafted patterns and their focus on binary relations. It utilizes SpanBERT to capture relational patterns from text data directly and introduces entity embedding vectors to create distinct representations of entities within sentences. These vectors enhance the proposed system's understanding of the entities within the input sentence, leading to more accurate relation extraction. Notably, the proposed system in the present study achieves an F1-score of 89.79 and 92.67 on the LSOIE-wiki and OpenIE4 datasets, outperforming the best existing models by over 12% and 10%, respectively. These results highlight the effectiveness of the proposed approach in addressing the challenges of open n-ary relation extraction. © 2024 Elsevier Inc.
Cybernetics and Systems (01969722) 56(1)pp. 1-20
The Controllability on temporal complex networks is one of the most important challenges among researchers in this field. The primary purpose of network controllability is to apply inputs by selecting minimum driver nodes set (MDS) to the network components to move the network from an initial state to a final state in a limited time. The most important challenges in the controllability of temporal networks can be mentioned the high complexity of the control algorithms used in these methods as well as the high data overhead of temporal network representation models such as the layered model. In this paper, centrality measures are used as the most important characteristics of networks for network controllability. For this purpose, centrality measures have been redefined based on temporal networks and a new controllability method has been proposed based on temporal centrality measures. Then, these properties are used for selecting the minimal driver nodes set, in such a way that the network can be fully controlled using these nodes. The experimental results demonstrate that by using temporal centrality measures the execution speed of control processes is improved (57% improvement) and the overhead is not increased and also the control process has led to the same length of MDS as other conventional controllability methods, it has even been better in some cases. © 2022 Taylor & Francis Group, LLC.
Neurocomputing (09252312) 616
Open Information Extraction (Open IE) is the task of identifying structured and machine-readable information from natural language text within an open domain context. This research area has gained significant importance in the field of natural language processing (NLP), attracting considerable attention for its potential to extract valuable information from unstructured textual data. Previous investigations heavily relied on manual extraction patterns and various NLP tools. While these methods often produce errors that accumulate and propagate throughout the systems, ultimately affecting the accuracy of the results. Moreover, recent Open IE studies have focused on extracting binary relations involving two entities. However, these binary approaches occasionally lead to the omission of essential information in the text, preventing a deeper comprehension of the content. This limitation arises from the fact that real-world relations often involve multiple entities, but binary approaches may oversimplify these relations and miss additional details crucial for a thorough understanding of text. To address these challenges, our study introduces an innovative system called “Open N-ary Information EXtraction (ONIEX).” This system incorporates two novel techniques: multihead relation attention mechanism and relation embedding. Multihead relation attention, in combination with relation embedding, enables the system to focus on relations extracted through the SpanBERT model and accurately identify associated entities for each relation. The ONIEX system's superior performance is substantiated through extensive experiments conducted on the OpenIE4 and LSOIE datasets, benchmark datasets for Open n-ary Information Extraction (Open n-ary IE). The results demonstrate the superiority of the ONIEX system over the existing state-of-the-art systems. © 2024 Elsevier B.V.
Expert Systems with Applications (09574174) 275
Entity search websites enable consumers to purchase products and entities. These websites commonly provide predefined search filters and contain entity information and user-created reviews. Consumers commonly exploit filters to eliminate irrelevant entities. Consequently, they explore reviews about the purified entities to identify an entity centered on their preferences. Inspired by this consumer-centric process, this paper introduces a mechanized model for entity search. In the presented model, consumer preferences are considered as queries comprising configurations for predefined filters and a textual part containing more explanations and constrains. Previous works on entity retrieval have not employed capabilities of structured entity information, and complete potential of insights in user reviews. In contrast, the presented model, considering this limitations, exploits structured entity information of predefined_slots to filter out many irrelevant entities. Consequently, it ranks entities by comparing text_part of consumer preferences with the entire core content of user reviews, indicating the order of corresponding entities centered on consumer preferences. The entity search model presented in this paper employs cutting-edge natural language processing (NLP) methods. Moreover, this paper introduces a model to curate entity search dataset, considering structured entity information and objective information in user reviews and exploit it to create a dataset by extracting and labeling restaurant information from Tripadvisor. The created entity search model incorporates a monoBERT-based text_ranker, fine-tuned employing the created training dataset, and evaluations indicate perceptible improvements in mean reciprocal rank (MRR), mean average precision (MAP), and normalized discounted cumulative gain (nDCG). © 2025 Elsevier Ltd
PLoS ONE (19326203) 20(5 May)
The extraction of subjective comparative relations is essential in the field of question answering systems, playing a crucial role in accurately interpreting and addressing complex questions. To tackle this challenge, we propose the SCQRE model, specifically designed to extract subjective comparative relations from questions by focusing on entities, aspects, constraints, and preferences. Our approach leverages multi-task learning, the Natural Language Inference (NLI) paradigm, and a specialized adapter integrated into RoBERTa_base_go_emotions to enhance performance in Element Extraction (EE), Compared Elements Identification (CEI), and Comparative Preference Classification (CPC). Key innovations include handling X- and XOR-type preferences, capturing implicit comparative nuances, and the robust extraction of constraints often neglected in existing models. We also introduce the Smartphone-SCQRE dataset, along with another domain-specific dataset, BrandsCompSent-19-SCQRE, both structured as subjective comparative questions. Experimental results demonstrate that our model outperforms existing approaches across multiple question-level and sentence-level datasets and surpasses recent language models, such as GPT-3.5-turbo-0613, Llama-2-70b-chat, and Qwen-1.5-7B-Chat, showcasing its effectiveness in question-based comparative relation extraction. © 2025 Babaali et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Pervasive and Mobile Computing (15741192) 105
In recent years, one type of complex network called the Social Internet of Things (SIoT) has attracted the attention of researchers. Controllability is one of the important problems in complex networks and it has essential applications in social, biological, and technical networks. Applying this problem can also play an important role in the control of social smart cities, but it has not yet been defined as a specific problem on SIoT, and no solution has been provided for it. This paper addresses the controllability problem of the temporal SIoT network. In this regard, first, a definition for the temporal SIoT network is provided. Then, the unique relationships of this network are defined and modeled formally. In the following, the Controllability problem is applied to the temporal SIoT network (CSIoT) to identify the Minimum Driver nodes Set (MDS). Then proposed CSIoT is compared with the state-of-the-art methods for performance analysis. In the obtained results, the heterogeneity (different types, brands, and models) has been investigated. Also, 69.80 % of the SIoT sub-graphs nodes have been identified as critical driver nodes in 152 different sets. The proposed controllability deals with network control in a distributed manner. © 2024 Elsevier B.V.
Question-answering systems, characterized by their fundamental functions of question classification, information retrieval, and answer selection, demand refinement to enhance precision in retrieving exact answers. Question classification, a cornerstone task, anticipates the probable answer to a posed query. However, the performance of question classification algorithms is hampered, particularly in agglutinative languages with complex morphology like Persian, where linguistic resources are limited. In this study, we propose a novel multi-layer Long-short-term memory (LSTM) Attention Convolutional Neural Network (CNN) (LACNN) classifier, tailored to extract pertinent information from Persian language contexts. Notably, this model operates autonomously, obviating the need for prior knowledge or external features. Moreover, we introduce UIMQC, the first medical question dataset in Persian, derived from the English GARD dataset. The inquiries within UIMQC are inherently intricate, often pertaining to rare diseases necessitating specialized diagnosis. Our experimental findings demonstrate a notable enhancement over baseline methods, with a 9% performance increase on the UTQC dataset, and achieving 67.08% accuracy on the UIMQC dataset. Consequently, we advocate for the adoption of the LACNN model in various morphological analysis tasks across low-resource languages, as in Question Answering systems it improves the performance for retrieving accurate answers to the users’ queries. ©2024 IEEE.
Applied Soft Computing (15684946) 156
Personalization of game difficulty is a critical task in leveraging artificial intelligence (AI) technologies to enhance player engagement in virtual worlds like metaverse. One of the key challenges in this area is developing methods for assessing a player's perception of game difficulty. This information can be used to dynamically adjust the game difficulty to match the player's skill level and preferences, which can improve the player's experience and engagement. The existing approaches have limitations such as relying on costly external devices, requiring time-consuming feedback or questionnaires, and being specific to certain game genres and narratives. In this paper, we propose a new method called ChatDDA for evaluating a player's perception of game difficulty by analyzing the content of their chat messages. Our method uses a pre-trained language model to extract semantic features from the chat messages, which are then used to train a feed-forward neural network to predict the player's level of hopefulness or despair about succeeding in the game. Three pre-trained language models—BERT, RoBERTa, and Twitter-roBERTa—are fine-tuned on a purpose-built dataset of player chat messages of the popular multiplayer online game PlayerUnknown's Battlegrounds (PUBG) tagged as expressing optimism or pessimism regarding game success. The results showed that our method can accurately predict a player's perception of game difficulty, with an accuracy of 0.953 on the test dataset of player chat messages. This suggests that our method has the potential to enhance player engagement and immersion within the game, ultimately leading to more satisfying and enjoyable metaverse experiences. © 2024 Elsevier B.V.
Domain model stands as a crucial part of software engineering, emerging from collaborative team efforts. Domain modeling involves creating a conceptual representation of a specific problem domain to determine the concepts and relationships. A pivotal step in domain modeling is reviewing the domain model to identify errors and abnormalities. Typically, software engineers engage in a manual review of the domain model diagrams for refinement purposes. However, the process of detecting errors and abnormalities can be time-consuming and error prone. Furthermore, it relies heavily on the expertise of software engineers. A primary concern in domain modeling is the repetition of concepts, which often occurs due to the involvement of multiple engineers. Recently, AI techniques have exhibited remarkable ability in modeling domain. This paper proposes an approach, called ARDEMIS, for equivalence checking that aims to detect semantically similar concepts within domain models. ARDEMIS utilizes the combination of pre-trained model and dictionary to identify equivalent elements. We assess our approach using a real-world case study in the transportation domain. Our findings reveal the ability of ARDEMIS to identify potential equivalent elements. © 2024 IEEE.
Simulation Modelling Practice and Theory (1569190X) 131
Emergency response environments in which the crisis occurs include many homogenous and heterogeneous entities, e.g., organizations and persons. These entities need to collaborate to address a crisis. However, most of the time, the organizations that are involved in a crisis suffer from communication and coordination problems. This may cause unfair resource distribution. To mitigate the problem, there is a pressing need to construct models of emergency response environments and simulate the coordination strategies proactively, before the onset of any crisis. This paper highlights the complexity of the system formed by interacting organizations during a crisis and the difficulty in programming for the simulation of such complex systems. To cope with this issue, this paper presents an executable domain-specific modeling language called CoorERE. The language facilitates modeling emergency response environments for the simulation of one of the most common coordination strategies, i.e., auction-based coordination strategies in crisis response. CoorERE is used for the graphical modeling of emergency response environments and simulation of both inter and intra-organizational auction-based coordination strategies. The paper presents two case studies with different complexities to show the applicability of CoorERE. An empirical study is conducted to compare CoorERE with MDD4ABMS, which is one of the most recent tools for modeling and simulation of multi-agent systems. The evaluation includes both objective and subjective assessments. The objective assessment considers the development effort required for modeling a complete case study, while the subjective assessment compares the tools in terms of satisfying modeling method requirements and quality metrics such as usability, productivity, and expressiveness. The results of the evaluation indicate that CoorERE reduces the development effort by about 47%. In terms of quality metrics, CoorERE obtained higher scores than MDD4ABMS for all the measurements considered, indicating its superior usability, productivity, and expressiveness. © 2023
Expert Systems with Applications (09574174) 252
The detection of mentioned aspects in product reviews is one of the significant and complex tasks in opinion mining. Recently, contextual-based approaches have significantly improved the accuracy of aspect extraction over non-contextual embeddings. However, these approaches are often computationally expensive and time-consuming; thus, applying such heavy models with insufficient resources and within runtime systems is impractical in many realistic scenarios. The present investigation sought an efficient, practical deep-learning-based model that relies on the complementary power of various existing non-contextual embeddings. In this regard, two morphology-based (character and FastText) and two syntax-based (POS and extended dependency skip-gram) embeddings were used alongside a base word embedding (GloVe) to form an enriched word representation layer. The presented model was integrated into the proposed network architecture (extended BiGRU). Finally, two novel post-processing rules were applied to refine the errors in the model's predictions. The proposed model achieved F-scores of 0.86, 0.91, 0.79, and 0.80 for the SemEval 2014 laptop domain and the SemEval 2015–2016 restaurant domain, respectively. Furthermore, the results were validated by comparing the computational and temporal efficiency of the proposed model with seven BERT-family transformers through statistical tests. © 2024 Elsevier Ltd
PLoS ONE (19326203) 19(5 May)
In the domain of question subjectivity classification, there exists a need for detailed datasets that can foster advancements in Automatic Subjective Question Answering (ASQA) systems. Addressing the prevailing research gaps, this paper introduces the Fine-Grained Question Subjectivity Dataset (FQSD), which comprises 10,000 questions. The dataset distinguishes between subjective and objective questions and offers additional categorizations such as Subjective-types (Target, Attitude, Reason, Yes/No, None) and Comparison-form (Single, Comparative). Annotation reliability was confirmed via robust evaluation techniques, yielding a Fleiss's Kappa score of 0.76 and Pearson correlation values up to 0.80 among three annotators. We benchmarked FQSD against existing datasets such as (Yu, Zha, and Chua 2012), SubjQA (Bjerva 2020), and ConvEx-DS (Hernandez-Bocanegra 2021). Our dataset excelled in scale, linguistic diversity, and syntactic complexity, establishing a new standard for future research. We employed visual methodologies to provide a nuanced understanding of the dataset and its classes. Utilizing transformer-based models like BERT, XLNET, and RoBERTa for validation, RoBERTa achieved an outstanding F1-score of 97%, confirming the dataset's efficacy for the advanced subjectivity classification task. Furthermore, we utilized Local Interpretable Model-agnostic Explanations (LIME) to elucidate model decision-making, ensuring transparent and reliable model predictions in subjectivity classification tasks. Copyright: © 2024 Babaali et al.
Multimedia Tools and Applications (13807501) 83(10)pp. 31049-31079
The game industry is witnessing a significant trend of players toward massively multiplayer online games (MMO). Players are keen on forming teams and cooperating/competing in these games. Real-time measurement of players’ performance is one of the subjects of researchers’ attention to dynamically adjust the game difficulty and immerse players in the game. However, our extensive studies show that real-time measuring of teams’ skill levels has received much less attention. In this paper, a general real-time method called DeepSkill is proposed to measure the MMOs teams’ skills directly using players’ gameplay raw low-level data. The proposed method, which is based on the evidence-centered assessment design, was tested under six different configurations using popular machine learning techniques, including deep neural network (DNN), extreme gradient boosting (XGBoost), light gradient boosting machine (LightGBM), CatBoost, random forest (RF), and linear support vector regression (LinearSVR). According to the results, the proposed method provides accurate skill estimations and expertise level classifications. Specifically, Deepskill’s DNN-based evidence model provided the lowest mean absolute error of 0.09 in team skill estimation. Additionally, the proposed method achieved an accuracy of 0.973 in classifying the teams’ expertise level for the expert-novice classification task. Furthermore, a cost-effectiveness analysis was performed on the two top-performing evidence models. The LightGBM-based evidence model yielded the best results in both training and prediction phases in terms of low resource consumption alongside considerable accuracy. © The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2023.
Computers in Human Behavior (07475632) 151
Assessing players’ learning experiences in a proper manner is a fundamental aspect of successful game-based learning programs. One notable characteristic of these programs is stealth assessment, which involves integrating formative assessment into the learning environment without disrupting the learning process. In multiplayer online games (MOGs), the in-game online chat system is a commonly used tool that enables players to communicate through text or voice messages during gameplay. However, there is a lack of specific research on incorporating players’ in-game chat content for computational learning experience assessment, which could enhance the validity of stealth assessment. This study proposes a stealth assessment method based on natural language processing to highlight the significance of players’ in-game chat data in estimating learners’ skills in MOGs. A natural language processing model is developed using a distilled version of the Google BERT pre-trained model. The evaluations demonstrate that the proposed method accurately estimates a player's skill level by analyzing a few chat messages from the player. This method has the potential to make a profound impact on the field of game-based learning by enabling more precise assessment and supporting the design of tailored interventions and adaptive learning systems. This study pioneers computational skill assessment through chats in MOGs, opening up new opportunities for future investigations in skill assessment and having the potential to transform the field of game-based learning. © 2023 Elsevier Ltd
Engineering Applications of Artificial Intelligence (09521976) 126
COVID-19 is a huge global challenge, on which numerous people and resources all over the world have been focused over the last three years to slow down its spread. A screening scale for the early detection of COVID-19 cases is highly important to diagnose patients before they transmit the disease to others. Using the data collected from Sept. 2020 to Nov. 2020 from 12,123 employees of a company including a weekly self-report questionnaire and information on their demographics, chronic disease, work, and commute, we developed a short, optimized screening scale for the early detection of COVID-19. To develop the scale, we have modified an interpretable machine learning (ML) method which uses integer programming techniques, and consequently reduced its computational cost significantly and made it robust to erroneous data and protective of sensitive information without loss of performance. The developed scale is highly interpretable and easily scored to the extent that it does not require a computer or even a calculator. However, we showed that the operating characteristics of the screening scale on the test dataset are excellent (area under the ROC curve = 82% and area under the Precision–Recall curve = 27%) and comparable to other well-known ML models such as penalized logistic regression and random forest when predicting positive PCR test results within 7 days from the date the self-report questionnaire was filled out. From Dec. 2020 to Nov. 2021, the developed scale has been used in the company weekly to monitor and assess its employees. Although different variants of COVID-19 have emerged since the development of the scale, the results show that the scale has performed well in the early detection of COVID-19, which helped the company take appropriate actions to slow down the spread of COVID-19 among its employees successfully. We expect our findings in this paper to help researchers have a better understanding of COVID-19 and control the spread of the disease. © 2023 Elsevier Ltd
Expert Systems with Applications (09574174) 213
In recent years, controllability on complex networks has become one of the most important issues among researchers. This study addresses the problem of controllability on an event-based complex network using events and their resulting dynamics to fully control the network. A particular type of event-based complex network, named event-based social networks (EBSNs), has been selected as a case study. In these networks, the communications between users are established by different event streams. A new control method, called Event Stream Controllability, is provided that uses the concept of maximum controllable subspace and maintains the data required for controlling the network using a tree structure. The experimental results demonstrate that the proposed method fully controls the network with a small number of control nodes (13.86%). In addition, it has been compared with the structural controllability based on the layer model. The results demonstrate that the proposed method outperforms the structural controllability method by 39.85%, 39.42%, and 34.98% increases in the number of driver nodes, runtime, and overload, respectively. Finally, the results show that the hub nodes (2%) and the organizer nodes (0.75%) are presented in the set of driver nodes, indicating that the proposed method is highly robust. © 2022 Elsevier Ltd
Expert Systems with Applications (09574174) 229
Question Classification is the very first and pivotal step in Question Answering(QA) systems. It maps a given question into a predefined class. Classification of Consumer Health Questions contributes to answering medical queries of non-expert users. It plays a principal role in the performance of Consumer Health Question Answering systems. There are several approaches to the classification of questions. However, they do not consider the nature of Consumer Health Questions. In this study, we perform experiments focusing on the question representation methods to classify Consumer Health Questions accurately. These experiments provide an intuition of practical considerations for Consumer Health Question classification. The questions were represented through certain manually-designed features, word embedding, sentence embedding, and finetuning-based techniques. We create question classifiers that involve designing models for representing and categorizing questions. Based on the result, the fine-tuned Bidirectional Encoder Representations from Transformers(BERT)-large-based model with an accuracy of 86.00% on Genetic and Rare Diseases (GARD) and 80.40% on Yahoo! Answers-based datasets outperform previous works. In addition, the method based on A Robustly Optimized BERT Pretraining Approach(RoBERTa)-large model with 86.20% accuracy surpasses fine-tuned BERT-large model on the GARD dataset. Examining the outcomes of the models gives insight into decisive considerations for representing and classifying Consumer Health Questions. © 2023 Elsevier Ltd
This paper presents the system developed by the Sartipi-Sedighin team for SemEval 2023 Task 2, which is a shared task focused on multilingual complex named entity recognition (NER), or MultiCoNER II. The goal of this task is to identify and classify complex named entities (NEs) in text across multiple languages. To tackle the MultiCoNER II task, we leveraged pre-trained language models (PLMs) fine-tuned for each language included in the dataset. In addition, we also applied a data augmentation technique to increase the amount of training data available to our models. Specifically, we searched for relevant NEs that already existed in the training data within Wikipedia, and we added new instances of these entities to our training corpus. Our team achieved an overall F1 score of 61.25% in the English track and 71.79% in the multilingual track across all 13 tracks of the shared task that we submitted to. © 2023 Association for Computational Linguistics.
With the advance of Model-Driven Engineering (MDE), number of generated models has grown exponentially. From the one hand, to provide support for software development using MDE, we need to store various models in the model repositories. From the other hand, extracting a particular model of interest from model repositories is a major challenge. One of the solutions is to use topic tracking in model mining. This provides a rapid and robust exploration that aligns well with the process of finding desired models in complex and large model repositories. In this paper, we propose TrackMine, a genetic algorithm based on topic tracking, to effectively sort the arrangement of models in a repository. This can help modelers realize the tracking, usage, and evolution of the models. TrackMine creates an optimized list of models by rearranging the order of models using user-defined similarity metrics. We demonstrate the applicability of our approach through a proof of concept implementation and evaluate the benefits of the presented algorithm using three different datasets. The experimental results demonstrate the practicality and suitability of our approach for model mining. © 2023 IEEE.
Journal of Supercomputing (15730484) 78(1)pp. 596-615
Today, social networks have created a wide variety of relationships between users. Friendships on Facebook and trust in the Epinions network are examples of these relationships. Most social media research has often focused on positive interpersonal relationships, such as friendships. However, in many real-world applications, there are also networks of negative relationships whose communication between users is either distrustful or hostile in nature. Such networks are called signed networks. In this work, sign prediction is made based on existing links between nodes. However, in real signed networks, links between nodes are usually sparse and sometimes absent. Therefore, existing methods are not appropriate to address the challenges of accurate sign prediction. To address the sparsity problem, this work aims to propose a method to predict the sign of positive and negative links based on clustering and collaborative filtering methods. Network clustering is done in such a way that the number of negative links between the clusters and the number of positive links within the clusters are as large as possible. As a result, the clusters are as close as possible to social balance. The main contribution of this work is using clustering and collaborative filtering methods, as well as proposing a new similarity criterion, to overcome the data sparseness problem and predict the unknown sign of links. Evaluations on the Epinions network have shown that the prediction accuracy of the proposed method has improved by 8% compared to previous studies. © 2021, The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature.
Aghaee, Z. ,
Ghasemi, M.M. ,
Beni, H.A. ,
Bouyer, A. ,
Fatemi, A. Computing (0010485X) 103(11)pp. 2437-2477
The different communications of users in social networks play a key role in effect to each other. The effect is important when they can achieve their goals through different communications. Studying the effect of specific users on other users has been modeled on the influence maximization problem on social networks. To solve this problem, different algorithms have been proposed that each of which has attempted to improve the influence spread and running time than other algorithms. Due to the lack of a review of the meta-heuristic algorithms for the influence maximization problem so far, in this paper, we first perform a comprehensive categorize of the presented algorithms for this problem. Then according to the efficient results and significant progress of the meta-heuristic algorithms over the last few years, we describe the comparison, advantages, and disadvantages of these algorithms. © 2021, The Author(s), under exclusive licence to Springer-Verlag GmbH Austria, part of Springer Nature.
Multimedia Tools and Applications (13807501) 80(9)pp. 13559-13574
In emotion-aware music recommender systems, the user’s current emotion is identified and considered in recommending music to him. We have two motivations to extend the existing systems: (1) to the best of our knowledge, the current systems first estimate the user’s emotions and then suggest music based on it. Therefore, the emotion estimation error affects the recommendation accuracy. (2) Studies show that the pattern of users’ interactions with input devices can reflect their emotions. However, these patterns have not been used yet in emotion-aware music recommender systems. In this study, a music recommender system is proposed to suggest music based on users’ keystrokes and mouse clicks patterns. Unlike the previous ones, the proposed system maps these patterns directly to the user’s favorite music, without labeling its current emotion. The results show that even though this system does not use any additional device, it is highly accurate compared to previous methods. © 2021, The Author(s), under exclusive licence to Springer Science+Business Media, LLC part of Springer Nature.
Recent advances in the field of Question Answering (QA) have improved state-of-the-art results. Due to the availability of rich English training datasets for this task, most results reported are for this language. However, due to the lack of Persian datasets, less research has been done for the latter language therefore the results are hard to compare. In the present work, we introduce the Persian Question Answering Dataset (ParSQuAD) translated from the well-known SQuAD 2.0 dataset. Our dataset comes in two versions depending on whether it has been manually or automatically corrected. The result is the first large-scale QA training resource for Persian. We train three baseline models, one of which, achieves an F1 score of 56.66% and an exact match ratio of 52.86% on the test set with the first version and an F1 score of 70.84 % and an exact match ratio of 67.73% with the second version. © 2021 IEEE.
Distributed and Parallel Databases (09268782) 38(1)
Relations among data entities in most big data sets can be modeled by a big graph. Implementation and execution of algorithms related to the structure of big graphs is very important in different fields. Because of the inherently high volume of big graphs, their calculations should be performed in a distributed manner. Some distributed systems based on vertex-centric model have been introduced for big graph calculations in recent years. The performance of these systems in terms of run time depends on the partitioning and distribution of the graph. Therefore, the graph partitioning is a major concern in this field. This paper concentrates on big graph partitioning approaches for distribution of graphs in vertex-centric systems. This briefly discusses vertex-centric systems and formulates different models of graph partitioning problem. Then, a review of recent methods of big graph partitioning for these systems is shown. Most recent methods of big graph partitioning for vertex centric systems can be categorized into three classes: (i) stream-based methods that see vertices or edges of the graph in a stream and partition them, (ii) distributed methods that partition vertices or edges in a distributed manner, and (iii) dynamic methods that change partitions during the execution of algorithms to obtain better performance. This study compares the properties of different approaches in each class and briefly reviews methods that are not in these categories. This comparison indicates that The streaming methods are good choices for initial load of the graph in Vertex-centric systems. The distributed and dynamic methods are appropriate for long-running applications. © 2019, Springer Science+Business Media, LLC, part of Springer Nature.
IET Information Security (17518709) 14(3)pp. 332-341
One of the main challenges of centralised social networks is having a central provider that stores the data which imposes some limitations to preserve the privacy of users' data. However, one of the decentralised architectures is peer-to-peer network that every user takes the responsibility of storing and managing his/her data. Although the privacy of data is increased in these networks, authorised friends must have access to the shared data when the user is not online in the network. For this purpose, the user selects some friends and copies his/her data in their space. On the other hand, the amount of used space and the total number of replicas must be reduced as much as possible. In this study, the authors provide some solutions to reduce the amount of used space and the total number of replicas to increase data availability. In this way, they segment the user's data and consider the stability of copy-location, i.e. the selected friends who have a copy of the user's data. The performance evaluation of the proposed methods shows that they considerably reduce the amount of used space as well as the total number of replicas in comparison to other approaches. © The Institution of Engineering and Technology 2020.
Turkish Journal Of Electrical Engineering And Computer Sciences (13000632) 28(3)pp. 1474-1490
RDF-based question answering systems give users the capability of natural language querying over RDF data. In order to respond to natural language questions, it is necessary that the main concept of the question be interpreted correctly, and then it is mapped to RDF data. A natural language question includes entities, classes, and implicit and explicit relationships. In this article, by focusing on identification and mapping of implicit relationships in the question (in addition to the explicit relationships), the mapping step has been improved. In the proposed solution (IRQA), entities and implicit/explicit relationships are identified by means of the provided rules and then will be presented as a graph. In the next phase, according to the determined priority of graph nodes, unknown nodes and their adjacent relationships are mapped to their peers in the RDF dataset. The proposed solution is evaluated on the QALD-3 test set. The results prove that the proposed work has improved performance in mapping implicit relationships and reveals higher precision and F-measure values. © 2020 Turkiye Klinikleri. All rights reserved.
Hoseindoost, S. ,
Adamzadeh, T. ,
Zamani, B. ,
Fatemi, A. Software and Systems Modeling (16191366) 18(3)pp. 1985-2012
In emergency response environments, variant entities with specific behaviors and interaction between them form a complex system that can be well modeled by multi-agent systems. To build such complex systems, instead of writing the code from scratch, one can follow the model-driven development approach, which aims to generate software from design models automatically. To achieve this goal, two important prerequisites are: a domain-specific modeling language for designing an emergency response environment model, and transformation programs for automatic code generation from a model. In addition, for modeling with the language, a modeling tool is required, and for executing the generated code there is a need to a platform. In this paper, a model-driven framework for developing multi-agent systems in emergency response environments is provided which includes several items. A domain-specific modeling language as well as a modeling tool is developed for this domain. The language and the tool are called ERE-ML and ERE-ML Tool, respectively. Using the ERE-ML Tool, a designer can model an emergency response situation and then validate the model against the predefined constraints. Furthermore, several model to code transformations are defined for automatic multi-agent system code generation from an emergency response environment model. For executing the generated code, an extension of JAMDER platform is also provided. To evaluate our framework, several case studies including the Victorian bushfire disaster are modeled to show the ability of the framework in modeling real-world situations and automatic transformation of the model into the code. © 2017, Springer-Verlag GmbH Germany.
Khodabandeh shahraki, Z. ,
Fatemi, A. ,
Tabatabaee malazi, H. Information Processing and Management (03064573) 56(6)
The widespread popularity and worldwide application of social networks have raised interest in the analysis of content created on the networks. One such analytical application and aspect of social networks, including Twitter, is identifying the location of various political and social events, natural disasters and so on. The present study focuses on the localization of traffic accidents. Outdated and inaccurate information in user profiles, the absence of location data in tweet texts, and the limited number of geotagged posts are among the challenges tackled by location estimation. Adopting the Dempster–Shafer Evidence Theory, the present study estimates the location of accidents using a combination of user profiles, tweet texts, and the place attachments in tweets. The results indicate improved performance regarding error distance and average error distance compared to previously developed methods. The proposed method in this study resulted in a reduced error distance of 26%. © 2019 Elsevier Ltd
In recent years, question and answer systems and information retrieval have been widely used by web users. The purpose of these systems is to find answers to users' questions. These systems consist of several components that the most essential of which is the Answer Selection, which finds the most relevant answer. In related works, the proposed models used lexical features to measure the similarity of sentences, but in recent works, the line of research has changed. They used deep neural networks. In the deep neural networks, early, recurrent neural networks were used due to the sequencing structure of the text, but in state of the art works, convolutional neural networks are used. We represent a new method based on deep neural network algorithms in this research. This method attempts to find the correct answer to a given question from the pool of responses. Our proposed method uses wide convolution instead of narrow convolution, concatenates sparse features vector into feature vector and uses dropout in order to rank candidate answers of the user's question semantically. The results show a 1.01% improvement at the MAP and a 0.2% improvement at the MRR metrics than the best previous model. The experiments show using context-sensitive interactions between input sentences is useful for finding the best answer. © 2019 IEEE.
Knowledge and Information Systems (02191377) 61(2)pp. 847-871
Relations among data items can be modeled with graphs in most of big data sets such as social networks’ data. This modeling creates big graphs with many vertices and edges. Balanced k-way graph partitioning is a common problem with big graphs. It has many applications in several fields. There are many approximate solutions for this problem; however, most of them do not have enough scalability for big graph partitioning and cannot be executed in a distributed manner. Vertex-centric model has been introduced recently as a scalable distributed processing method for big graphs. There are a few methods for graph partitioning based on this model. Existing approaches only consider one-step neighbors of vertices for graph partitioning and do not consider neighbors with higher steps. In this paper, a distributed method is introduced based on vertex-centric model for balanced k-way graph partitioning. This method applies the personalized PageRank vectors of vertices and partitions to decide how vertices are joined partitions. This method has been implemented in the Giraph system. The proposed method has been evaluated with several synthetic and real graphs. Experimental results have shown that this method has scalability for partitioning big graphs. It was also found that this method produces partitions with higher quality compared to the state-of-the-art stream-based methods and distributed methods based on vertex-centric programming model. Its result is close to the results of Metis method. © 2019, Springer-Verlag London Ltd., part of Springer Nature.
Electronic Commerce Research (13895753) 18(4)pp. 813-836
In collaborative filtering recommender systems, items recommended to an active user are selected based on the interests of users similar to him/her. Collaborative filtering systems suffer from the ‘sparsity’ and ‘new user’ problems. The former refers to the insufficiency of data about users’ preferences and the latter addresses the lack of enough information about the new-coming user. Clustering users is an effective way to improve the performance of collaborative filtering systems in facing the aforementioned problems. In previous studies, users were clustered based on characteristics such as ratings given by them as well as their age, gender, occupation, and geographical location. On the other hand, studies show that there is a significant relationship between users’ personality traits and their interests. To alleviate the sparsity and new user problems, this paper presents a new collaborative filtering system in which users are clustered based on their ‘personality traits’. In the proposed method, the personality of each user is described according to the big-5 personality model and users with similar personality are placed in the same cluster using K-means algorithm. The unknown ratings of the sparse user-item matrix are then estimated based on the clustered users, and recommendations are found for a new user according to a user-based approach which relays on the interests of the users with similar personality to him/her. In addition, for an existing user in the system, recommendations are offered in an item-based approach in which the similarity of items is estimated based on the ratings of users similar to him/her in personality. The proposed method is compared to some former collaborative filtering systems. The results demonstrate that in facing the data sparsity and new user problems, this method reduces the mean absolute error and improves the precision of the recommendations. © 2018, Springer Science+Business Media, LLC, part of Springer Nature.
Turkish Journal of Electrical Engineering and Computer Sciences (13036203) 21(4)pp. 1166-1181
This paper addresses coalition formation, based on agent capabilities, centered on task allocation in emergency-response environments (EREs). EREs are environments that need fast task completion as their main requirement. We propose a team-based organization model, based on an existing organization model for adaptive complex systems. The model has some key characteristics that are beneficial for EREs: agents act in dynamic, open domains; agents collaborate in completing group tasks; agents may have similar types of capabilities, but at difierent levels; tasks need different agent capabilities, at collective different levels; and agents are supervised in a partially decentralized manner. We formulate task allocation as a capability-based coalition-formation problem, propose a greedy myopic algorithm to form coalitions, and compare it with F-Max-Sum, another effcient myopic algorithm. Experiments in which utility is measured show that the capability-based approach outperforms the role-based one. The numerical experiments suggest that the proposed task allocation method is possibly scalable with growing numbers of agents. © Tübi̇tak..
Fatemi, A. ,
Zamanifar, K. ,
Bakhsh, N.N. ,
Askari, O. 2pp. 469-472
Proper organizational modelling is a challenging issue in complex cooperative multi-agent systems. In this paper, we propose a team-based multi-agent organizational model, based on the Schwaninger's model of intelligent human organizations. It provides an integrative framework to rapid task handling, the main effectiveness requirement in many applications. Adaptation via reorganization makes the model suitable for dynamic, uncertain environments. Fast initial team formation, greedy capability-based coalition formation, and using the nearest neighbours' resources improve utility compared to the identified hierarchical organizational models.
AIP Conference Proceedings (0094243X) 1117pp. 180-188
ACS algorithms have been used in solving NP-hard and optimization problems in recent years. ACS ant colonies are homogeneous, but natural colonies are not. In this paper, a new ACS algorithm is proposed. It uses heterogeneous ant colonies which are evolved using a new type of genetic algorithm. Experimental results obtained from solving TSP problem, show the superiority of proposed algorithm over classical ACS. © 2009 American Institute of Physics.
Many definitive and approximate methods have been so far proposed for the construction of an optimal binary search tree. One such method is the use of evolutionary algorithms with satisfactorily improved cost efficiencies. This paper will propose a new genetic algorithm for constructing a near optimal binary search tree. In this algorithm, a new greedy method is used for the crossover of chromosomes while a new way is also developed for inducing mutation in them. Practical results show a rapid and desirable convergence towards the near optimal solution. The use of a heuristic to create not so costly chromosomes as the first offspring, the greediness of the crossover, and the application of elitism in the selection of future generation chromosomes are the most important factors leading to near optimal solutions by the algorithm at desirably high speeds. Due to the practical results, increasing problem size does not cause any considerable difference between the solution obtained from the algorithm and exact solution.
Applied Mathematics and Computation (00963003) 190(2)pp. 1514-1525
Many definitive and approximate methods have been so far proposed for the construction of an optimal binary search tree. One such method is the use of evolutionary algorithms with satisfactorily improved cost efficiencies. This paper will propose a new genetic algorithm for making a near-optimal binary search tree. In this algorithm, a new greedy method is used for the crossover of chromosomes while a new way is also developed for inducing mutation in them. Practical results show a rapid and desirable convergence towards the near-optimal solution. The use of a heuristic to create not so costly chromosomes as the first offspring, the greediness of the crossover, and the application of elitism in the selection of future generation chromosomes are the most important factors leading to near-optimal solutions by the algorithm at desirably high speeds. Due to the practical results, increasing problem size does not cause any considerable difference between the solution obtained from the algorithm and exact solution. © 2007 Elsevier Inc. All rights reserved.
Task allocation, as an important issue in multi-agent systems (MAS), is defined as allocating the tasks to the agents such that maximum tasks are performed in minimum time. The vast range of application domains, such as scheduling, cooperation in crisis management, and project management, deal with the task allocation problem. Despite the plethora of algorithms that are proposed to solve this problem in different application domains, research on proposing a formalism for this problem is scarce. Such a formalism can be used as a way for better understanding and analyzing the behavior of real-world systems. In this paper, we propose a new formalism for specifying capability-based task allocation in MAS. The formalism can be used in different application domains to help domain experts better analyze and test their algorithms with more precision. To show the applicability of the formalism, we consider two algorithms as the case studies and formalize the inputs and outputs of these algorithms using the proposed formalism. The results indicate that our formalism is promising for specifying the capability-based task allocation in MAS at a proper level of abstraction. © 2021 IEEE.
Question Answering is a hot topic in artificial intelligence and has many real-world applications. This field aims at generating an answer to the user's question by analyzing a massive volume of text documents. Answer Selection is a significant part of a question answering system and attempts to extract the most relevant answers to the user's question from the candidate answers pool. Recently, researchers have attempted to resolve the answer selection task by using deep neural networks. They first employed the recurrent neural networks and then gradually migrated to convolutional neural networks. Nevertheless, the use of language models, which is implemented by deep neural networks, has recently been considered. In this research, the DistilBERT language model was employed as the language model. The outputs of the Question Analysis part and Expected Answer Extraction component are also applied with [CLS] token output as the final feature vector. This operation leads to improving the method performance. Several experiments are performed to evaluate the effectiveness of the proposed method, and the results are reported based on the MAP and MRR metrics. The results show that the MAP values of the proposed method improved by 0.6%, and the MRR metric is improved by 0.2%. The results of our research show that using a heavy language model does not guarantee a more reliable method for answer selection problem. It also shows that the use of particular words, such as Question Word and Expected Answer word, can improve the performance of the method. © 2020 IEEE.
A modeling language is a way to describe syntax, semantic, and constraints needed for creating models. Defining a Domain Specific Modeling Language (DSML) instead of suing a general-purpose one, increases the productivity of the developer as well as the quality of the resulted model. In this paper, we proposed a DSML for the Mitigation phase of Emergency Response Environments (EREs). We extended the TAO framework based on the TAO provided textual patterns. This paper also involves extending MAS-ML to support the modeling of EREs Mitigation phase. To evaluate this work, a case study is modeled with the proposed modeling language. Higher abstraction level, less effort, and faster development process are results of the proposed modeling language. © 2014 IEEE.
Daher, H. ,
Hoseindoost, S. ,
Zamani, B. ,
Fatemi, A. pp. 35-41
In case of a disaster, planning for pedestrian evacuation from buildings is a major issue since it threatens human lives. To cope with this problem, evacuation plans are developed to ensure efficient evacuation in minimum time. These plans can be very sophisticated according to the complexity of the evacuation environment. This advocates the use of architectures such as Multi-Agent Systems (MAS) to develop the evacuation plans before happening of a real accident. Since developing an evacuation plan using MAS requires considerable effort, finding more efficient approaches is still an open problem. This paper introduces a new approach, based on the model-driven principles, to support developing evacuation plans. The approach includes utilizing a graphical editor for designing evacuation models, automatic generation of the evacuation plan code, as well as running the generated code on a MAS platform. We evaluated our approach using a case study. The results show that our approach provides elevated speed, less effort, high abstraction level, and more flexibility and productivity in developing emergency evacuation plans. © 2020 IEEE.
Authorship Attribution (AA) is a task in which a disputed text is automatically assigned to an author chosen from a list of candidate authors. To this end, a model is trained on a dataset of textual documents with known authors, which can be considered as a multi-class single-label classification task. In this paper, we approach this task differently by extending information retrieval techniques to train an AA model. It is based on weighting the AARR technique, presented in our previous study, to relax the value of term frequency. The efficiency of the proposed solution has been evaluated by conducting several experiments on six datasets. The results show the superiority of the proposed solution by improving the accuracy of IMDB, Gutenberg books, Poetry, Blogs, PAN2011, and Twitter datasets by 33%, 31%, 31%, 19%, 6%, and 1%, respectively, where the average improvement is 19.94% over all datasets. The best accuracy over these datasets is 88%, 82%, 67%, 90%, 65%, and 81% in the same respect. In addition, compared to the baseline system, the computation time of the proposed solution has been improved significantly (21.44X) by employing a dictionary-based indexing technique. © 2021 IEEE.
Hemmat, A. ,
Vadaei, K. ,
Shirian, M. ,
Heydari, M.H. ,
Fatemi, A.
This paper introduces an innovative approach to Retrieval-Augmented Generation (RAG) for video question answering (VideoQA) through the development of an adaptive chunking methodology and the creation of a bilingual educational dataset. Our proposed adaptive chunking technique, powered by CLIP embeddings and SSIM scores, identifies meaningful transitions in video content by segmenting educational videos into semantically coherent chunks. This methodology optimizes the processing of slide-based lectures, ensuring efficient integration of visual and textual modalities for downstream RAG tasks. To support this work, we gathered a bilingual dataset comprising Persian and English mid- to long-duration academic videos, curated to reflect diverse topics, teaching styles, and multilingual content. Each video is enriched with synthetic question-answer pairs designed to challenge pure large language models (LLMs) and underscore the necessity of retrieval-augmented systems. The evaluation compares our CLIP-SSIM-based chunking approach against conventional video slicing methods, demonstrating significant improvements across RAGAS metrics, including Answer Relevance, Context Relevance, and Faithfulness. Furthermore, our findings reveal that the multimodal image-text retrieval scenario achieves the best overall performance, emphasizing the importance of integrating complementary modalities. This research establishes a robust framework for video RAG pipelines, expanding the capabilities of multimodal AI systems for educational content analysis and retrieval. © 2025 IEEE.
Due to the growning use of social networks and the use of viral marketing in these networks, finding influential people to maximize information diffusion is considered. This problem is Influence Maximization Problem on social networks. The main goal of this Problem is to select a set of influential nodes to maximize the influence spread in a social network. Researchers in this field have proposed different algorithms, but finding the influential people in the shortest possible time is still a challenge that has attracted the attention of researchers. Therefore, in this paper, the IMPT-C algorithm is presented with a focus on graph pre-processing in order to reduce the search space based on community structure. The approach of this algorithm is to take advantage of the topological properties of the graph to identify influential nodes. The experiment results indicate that the IMPT-C algorithm has a great influence spread with low run time compared the state-of-the-art algorithms consist least 2.36% improve than PHG in term the influence spread. © 2021 IEEE.
The area of agent-oriented methodologies is maturing rapidly and the time has come to begin drawing together the work of various research groups with the aim of developing the next generation of agent-oriented software engineering methodologies. An important step is to understand the differences between the various key methodologies, and to understand each methodology's strengths, weaknesses, and domains of applicability. In this paper we perform an investigation upon user views, on four well-known methodologies. We extend Tropos, as the most complete one up on users view point, by providing a proper supportive tool for it. © 2006 IEEE.
Retrieval-Augmented Generation (RAG) has emerged as a widely adopted approach to mitigate the limitations of large language models (LLMs) in answering domain-specific questions. Previous research has predominantly focused on improving the accuracy and quality of retrieved data chunks to enhance the overall performance of the generation pipeline. However, despite ongoing advancements, the critical issue of retrieving irrelevant information—which can impair a model’s ability to utilize its internal knowledge effectively—has received minimal attention. In this work, we investigate the impact of retrieving irrelevant information in open-domain question answering, highlighting its significant detrimental effect on the quality of LLM outputs. To address this challenge, we propose the Context Awareness Gate (CAG) architecture, a novel mechanism that dynamically adjusts the LLM’s input prompt based on whether the user query necessitates external context retrieval. Additionally, we introduce the Vector Candidates method, a core mathematical component of CAG that is statistical, LLM-independent, and highly scalable. We further examine the distributions of relationships between contexts and questions, presenting a statistical analysis of these distributions. This analysis can be leveraged to enhance the context retrieval process in retrieval-augmented generation (RAG) systems. © 2024 IEEE.
This study focuses on the generation of Persian named entity datasets through the application of machine translation on English datasets. The generated datasets were evaluated by experimenting with one monolingual and one multilingual transformer model. Notably, the CoNLL 2003 dataset has achieved the highest F1 score of 85.11%. In contrast, the WNUT 2017 dataset yielded the lowest F1 score of 40.02%. The results of this study highlight the potential of machine translation in creating high-quality named entity recognition datasets for low-resource languages like Persian. The study compares the performance of these generated datasets with English named entity recognition systems and provides insights into the effectiveness of machine translation for this task. Additionally, this approach could be used to augment data in low-resource language or create noisy data to make named entity systems more robust and improve them. © 2023 IEEE.
This paper introduces an innovative approach using Retrieval-Augmented Generation (RAG) pipelines with Large Language Models (LLMs) to enhance information retrieval and query response systems for university-related question answering. By systematically extracting data from the university’s official website, primarily in Persian, and employing advanced prompt engineering techniques, we generate accurate and contextually relevant responses to user queries. We developed a comprehensive university benchmark, UniversityQuestionBench (UQB), to rigorously evaluate our system’s performance. UQB focuses on Persian-language data, assessing accuracy and reliability through various metrics and real-world scenarios. Our experimental results demonstrate significant improvements in the precision and relevance of generated responses, enhancing user experiences, and reducing the time required to obtain relevant answers. In summary, this paper presents a novel application of RAG pipelines and LLMs for Persian-language data retrieval, supported by a meticulously prepared university benchmark, offering valuable insights into advanced AI techniques for academic data retrieval and setting the stage for future research in this domain. © 2024 IEEE.
