Article
Computer Science, Information Systems
Jinlu Zhang, Jing He, Yiyi Zhou, Xiaoshuai Sun, Xiao Yu
Summary: In recent years, Question Answering (QA) systems have become popular for knowledge acquisition. However, the prevailing approach of matching question-answer pairs suffers from low precision and efficiency due to the ambiguity of natural language descriptions. To address this, we propose a hierarchical semantic matching-based QA approach, called HSM-QA. It involves query-question matching using a Siamese network for similarity calculation and query-answer matching using a pairwise algorithm and single-stream structure for relevance calculation. Experimental results demonstrate the superior performance and efficiency of our HSM-QA compared to other methods, along with the use of a new dataset generated from Chinese social media.
Article
Health Care Sciences & Services
Andre Esteva, Anuprit Kale, Romain Paulus, Kazuma Hashimoto, Wenpeng Yin, Dragomir Radev, Richard Socher
Summary: The CO-Search is a semantic, multi-stage search engine designed to handle complex queries over the COVID-19 literature, potentially aiding overburdened health workers in finding scientific answers and avoiding misinformation. It consists of a deep learning model and two keyword-based models that encode query-level meaning and assign relevance scores to each document based on how well they match the query.
NPJ DIGITAL MEDICINE
(2021)
Article
Engineering, Multidisciplinary
Betul Ay, Fatih Ertam, Guven Fidan, Galip Aydin
Summary: Text summarization is the process of reducing text size while preserving key points. This study focuses on abstract summarization and compares the performance of using the T5 method on a dataset collected from Turkish news sources. The results show successful precision, recall, and F measure values for the performance metrics used. The study presents a method for achieving successful Turkish text summarization results and provides the original dataset for further research.
ALEXANDRIA ENGINEERING JOURNAL
(2023)
Article
Computer Science, Information Systems
Negin Ghasemi, Ramin Fatourechi, Saeedeh Momtazi
Summary: The number of users with the appropriate knowledge to answer questions in community Q&A platforms is lower than those asking questions, making finding expert users crucial. This article presents a framework to identify and assign experts to questions by leveraging user relationships and semantic similarities. Experimentation on four Stack Exchange datasets demonstrates that incorporating community relations enhances expert finding models.
ACM TRANSACTIONS ON KNOWLEDGE DISCOVERY FROM DATA
(2021)
Article
Computer Science, Information Systems
Avaneesh Kumar Yadav, Rama Shankar Ranvijay, Rama Shankar Yadav, Ashish Kumar Maurya
Summary: With the increasing amount of online and offline text data, manual summarization of large documents is impractical and expensive. Graph-based text summarization techniques have been designed to provide well-prepared summaries, but they suffer from redundancy, information loss, and readability issues. To address these problems, a graph-based extractive summarization technique called TGETS has been proposed to extract essential information from a single document. The proposed approach achieves better scores than the existing PageRank (PR) method in terms of precision, recall, and F-1-score, showing its effectiveness in generating summaries.
MULTIMEDIA TOOLS AND APPLICATIONS
(2023)
Article
Computer Science, Artificial Intelligence
Chunxiao Fan, Wentong Chen, Yuexin Wu
Summary: Knowledge base question answering (KBQA) refers to combining the information in the knowledge base to obtain an answer for an objective question. Most existing methods involve adding hand-crafted features or constraints to improve model performance, resulting in complex KBQA systems with limited improvement. This work presents a novel method that avoids using hand-crafted features or constraints, transforming the problem into a text matching task. The Path Matching Model (PMM) is employed to match the question and a series of edges in the knowledge base (KB). On the WebQuestions benchmark, the method achieves a 3% improvement compared to the state-of-the-art approach.
KNOWLEDGE-BASED SYSTEMS
(2022)
Article
Geochemistry & Geophysics
Zixiao Zhang, Licheng Jiao, Lingling Li, Xu Liu, Puhua Chen, Fang Liu, Yuxuan Li, Zhicheng Guo
Summary: In this article, a novel method called spatial hierarchical reasoning network (SHRNet) is proposed to address the limitations of current methods in remote sensing visual question answering (RSVQA). The method enhances the visual-spatial reasoning capability and considers geospatial objects with large-scale differences and positional sensitive properties. Modeling and reasoning the relationships between entities are also explored for accurate answer predictions.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2023)
Article
Computer Science, Information Systems
Yongqi Li, Nan Yang, Liang Wang, Furu Wei, Wenjie Li
Summary: To address the issue of question ambiguity in conversation question answering, a generative retrieval method called GCoQA is proposed. GCoQA retrieves passages by generating their identifiers token-by-token via an encoder-decoder architecture, leading to significant improvements in passage retrieval and document retrieval compared to current methods.
INFORMATION PROCESSING & MANAGEMENT
(2023)
Article
Mathematics
Wael Etaiwi, Arafat Awajan
Summary: This study proposes a novel semantic graph embedding-based abstractive text summarization technique for the Arabic language, namely SemG-TS, and demonstrates its superiority through experiments.
Article
Computer Science, Artificial Intelligence
Tahani H. Alwaneen, Aqil M. Azmi, Hatim A. Aboalsamh, Erik Cambria, Amir Hussain
Summary: Question answering is a subfield of information retrieval that aims to answer questions posed in natural language. The development of Arabic question answering systems has been hindered by linguistic challenges and limited resources. Research in this area includes examining challenges, existing systems, techniques, and future directions for Arabic question answering systems.
ARTIFICIAL INTELLIGENCE REVIEW
(2022)
Review
Computer Science, Artificial Intelligence
Zahra Abbasiantaeb, Saeedeh Momtazi
Summary: Text-based question answering is a challenging task that aims to find short answers for user questions using information retrieval and deep learning techniques. This paper provides an overview of traditional IR and deep neural network models, introduces popular datasets, and presents a comparison of different techniques from the literature.
WILEY INTERDISCIPLINARY REVIEWS-DATA MINING AND KNOWLEDGE DISCOVERY
(2021)
Article
Computer Science, Information Systems
Minakshi Tomer, Manoj Kumar
Summary: This paper proposes a firefly algorithm-based approach for multi-document text summarization. The algorithm's performance is evaluated using datasets from DUC-2002, DUC-2003, and DUC-2004, and compared to other nature-inspired algorithms such as PSO and GA. The results show that the proposed algorithm outperforms the others in terms of summarization quality.
JOURNAL OF KING SAUD UNIVERSITY-COMPUTER AND INFORMATION SCIENCES
(2022)
Article
Computer Science, Interdisciplinary Applications
Tingting Zhang, Baozhen Lee, Qinghua Zhu, Xi Han, Ke Chen
Summary: A keyword extraction method based on a semantic hierarchical graph model is proposed in this paper. It fully takes into account the context and internal structure of keywords, and effectively reveals the hierarchical association between terms within the semantic graph by mining the deep hidden structure of feature terms. Experiments on released datasets show that the proposed method outperforms existing methods in terms of precision, recall, and F-measure.
Article
Computer Science, Information Systems
M. D. Rashad A. L. Hasan Rony, Debanjan Chaudhuri, Ricardo Usbeck, Jens Lehmann
Summary: This paper proposes an unsupervised KGQA system that utilizes pre-trained language models and tree-based algorithms. The system achieves significant improvements in entity and relation linking tasks, outperforming other state-of-the-art methods on the test set without training on the target dataset.
Article
Computer Science, Artificial Intelligence
Chunxiao Fan, Zhen Yan, Yuexin Wu, Bing Qian
Summary: Dense passage retrieval is a popular method in information retrieval, aiming to retrieve related articles from massive passages to answer questions. However, current pretrained language models suffer from ineffective semantic embedding, reducing accuracy. In addition, contrastive learning and Siamese models with independent parameters lead to diverged representation space and decreased generalization performance. To address these issues, the proposed span prompt dense passage retrieval (SPDPR) utilizes span mask prompt tuning and parameter sharing, achieving more efficient representation embeddings and countering the separation tendency between positive samples.
JOURNAL OF INTELLIGENT & FUZZY SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Katerina Papantoniou, Panagiotis Papadakos, Theodore Patkos, George Flouris, Ion Androutsopoulos, Dimitris Plexousakis
Summary: This research focuses on automatic deception detection in cross-cultural text, examining the impact of cultural differences on linguistic cues of deception from the individualism/collectivism dimension. The results indicate that the task is complex and demanding.
NATURAL LANGUAGE ENGINEERING
(2022)
Article
Genetics & Heredity
Nicholas Owen, Maria Toms, Rodrigo M. Young, Jonathan Eintracht, Hajrah Sarkar, Brian P. Brooks, Mariya Moosajee
Summary: In this study, new potential causative genes for ocular coloboma were identified using cross-species comparative meta-analysis. Through in silico analysis, in situ hybridization, gene knockdown, and rescue experiments, several differentially expressed genes were confirmed to be involved in the development of the optic fissure. Furthermore, novel pathogenic variants in four genes were identified in coloboma families. The findings demonstrate the utility of cross-species meta-analysis and provide insights into the genetic basis of ocular coloboma.
GENETICS IN MEDICINE
(2022)
Article
Multidisciplinary Sciences
Joel T. Gibson, Mary Huang, Marina Shenelli Croos Dabrera, Krushnam Shukla, Hansjorg Rothe, Pascale Hilbert, Constantinos Deltas, Helen Storey, Beata S. Lipska-Zietkiewicz, Melanie M. Y. Chan, Omid Sadeghi-Alavijeh, Daniel P. Gale, Agne Cerkauskaite, Judy Savige
Summary: This study examined the molecular characteristics of Gly substitutions in Alport syndrome and found that the stability of the substitution and its interaction with nearby residues determine the risk of haematuria, early onset kidney failure, and hearing loss in this inherited kidney disease.
SCIENTIFIC REPORTS
(2022)
Article
Multidisciplinary Sciences
Yannis Assael, Thea Sommerschield, Brendan Shillingford, Mahyar Bordbar, John Pavlopoulos, Marita Chatzipanagiotou, Ion Androutsopoulos, Jonathan Prag, Nando de Freitas
Summary: The study introduces Ithaca, a deep neural network for restoring, attributing, and dating ancient Greek inscriptions. The use of Ithaca significantly improves the accuracy of text restoration and attribution compared to historians working alone, contributing to the study of ancient history.
Article
Computer Science, Artificial Intelligence
John Pavlopoulos, Vasiliki Kougia, Ion Androutsopoulos, Dimitris Papamichail
Summary: Diagnostic captioning (DC) is the automatic generation of diagnostic text from medical images, assisting physicians in reducing errors and improving efficiency. With the advancements in deep learning, DC has gained attention and resulted in the development of various systems and datasets. This article provides an extensive overview of DC, including relevant datasets, evaluation measures, up-to-date systems, and proposed future directions.
KNOWLEDGE AND INFORMATION SYSTEMS
(2022)
Article
Computer Science, Artificial Intelligence
John Pavlopoulos, Aristidis Likas
Summary: Commonsense knowledge is often approximated by the fraction of annotators who classified an item as positive, which overlooks the polarization of opinions. We propose a novel measure, DFU, that estimates the extent of polarization and correlates well with human judgment. Applying DFU to pandemic-related tweets and toxic posts, we find that polarization occurs on different days for different states and is more likely among annotators from different countries. Furthermore, DFU can be used as an objective function to predict the potential for polarized opinions.
COGNITIVE COMPUTATION
(2023)
Article
Computer Science, Information Systems
Abdullah Fathi Ahmed, Mohamed Ahmed Sherif, Axel-Cyrille Ngonga Ngomo
Summary: This paper presents NELLIE, a pipeline architecture to build a chain of modules that address data augmentation challenges and ultimately build a fused knowledge graph out of Linked Open Data. NELLIE uses a two-phase linking approach to fuse each pair of knowledge graphs and improves the link prediction task's Hit@1 score by up to 94.44% compared to a naive approach.
Proceedings Paper
Computer Science, Artificial Intelligence
Caglar Demir, Axel-Cyrille Ngonga Ngomo
Summary: Concept learning is the process of learning description logic concepts from background knowledge and input examples. This study proposes a solution to the problem by formulating it as a multi-label classification problem and introducing a neural embedding model (NERO) that predicts F1 scores of selected concepts. By ranking the concepts based on predicted scores, a possible goal concept can be detected without excessive exploration.
ADVANCES IN INTELLIGENT DATA ANALYSIS XXI, IDA 2023
(2023)
Article
Management
Apostolos G. Katsafados, George N. Leledakis, Emmanouil G. Pyrgiotakis, Ion Androutsopoulos, Manos Fergadiotis
Summary: This paper investigates the role of textual information in a U.S. bank merger prediction task and finds that using textual information along with financial variables significantly improves the performance of the models, especially in predicting future bidders. These findings highlight the importance of textual information in a bank merger prediction task.
EUROPEAN JOURNAL OF OPERATIONAL RESEARCH
(2024)
Proceedings Paper
Computer Science, Artificial Intelligence
Manzoor Ali, Muhammad Saleem, Axel-Cyrille Ngonga Ngomo
Summary: In recent years, the development of relation extraction (RE) models has led to the proposal of several benchmark datasets for evaluating these models. However, these datasets do not allow for customized microbenchmarking according to user-specified criteria. This article presents the REBench framework, which enables the selection of customized relation samples from existing datasets in different domains for microbenchmarking. Evaluation of state-of-the-art RE systems using different benchmarking samples demonstrates the importance of specialized microbenchmarking in identifying limitations of RE models and their components.
SEMANTIC WEB - ISWC 2022
(2022)
Proceedings Paper
Computer Science, Artificial Intelligence
Alexander Biger, Lixi Conrads, Charlotte Behning, Muhammad Saleem, Axel-Cyrille Ngonga Ngomo
Summary: This paper introduces a method to reduce the memory footprint of hypertries, a tensor-based triple store indexing structure, in order to further improve query processing speed in RDF storage solutions. By eliminating duplicate nodes, compressing non-branching paths, and storing single-entry leaf nodes in their parent nodes, significant reductions in memory usage can be achieved.
SEMANTIC WEB - ISWC 2022
(2022)
Proceedings Paper
Computer Science, Artificial Intelligence
Umair Qudus, Michael Roeder, Muhammad Saleem, Axel-Cyrille Ngonga Ngomo
Summary: This paper investigates fact-checking approaches for knowledge graphs and introduces five main categories of approaches. Current methods have limitations such as manual feature engineering and exclusive use of knowledge graphs. To improve prediction performance, a hybrid approach is proposed that leverages the diversity of existing approaches within an ensemble learning setting.
SEMANTIC WEB - ISWC 2022
(2022)
Proceedings Paper
Computer Science, Artificial Intelligence
Hamada M. Zahera, Daniel Vollmers, Mohamed Ahmed Sherif, Axel-Cyrille Ngonga Ngomo
Summary: This paper proposes a multitask framework, MULTPAX, for keyphrase extraction using pre-trained language models and knowledge graphs. The experiments show that MULTPAX outperforms state-of-the-art baselines significantly.
SEMANTIC WEB - ISWC 2022
(2022)
Proceedings Paper
Computer Science, Artificial Intelligence
N'Dah Jean Kouagou, Stefan Heindorf, Caglar Demir, Axel-Cyrille Ngonga Ngomo
Summary: This paper investigates concept learning approaches based on refinement operators to address the efficiency issue in exploring solution spaces for complex learning problems. By predicting the length of target concepts, the search space can be pruned during concept learning. Experimental results suggest that recurrent neural network architectures perform the best in predicting concept length. The proposed CLIP algorithm, an extension of the CELOE algorithm, achieves significant improvements in both speed and the F-measure of concept learning.
SEMANTIC WEB, ESWC 2022
(2022)
Proceedings Paper
Computer Science, Cybernetics
Stefan Heindorf, Lukas Blubaum, Nick Dusterhus, Till Werner, Varun Nandkumar Golani, Caglar Demir, Axel-Cyrille Ngonga Ngomo
Summary: This paper proposes an evolutionary approach for learning concepts in knowledge graphs, which improves the initialization of the population and the support for data properties. The approach significantly outperforms existing techniques in structured machine learning tasks.
PROCEEDINGS OF THE ACM WEB CONFERENCE 2022 (WWW'22)
(2022)