Article
Multidisciplinary Sciences
Kenta Fukada, Michiko Seyama
Summary: Materials informatics (MI) is a valuable technique for designing chemical substances, but its application to layered structures is challenging. This study demonstrates that machine learning (ML) can be used to design multilayer films by extracting experimental procedures from chemical-coating articles. The ML approach connects scientific knowledge, enabling the prediction of untrained film structures. The results suggest that artificial intelligence (AI) can imitate research activity and serve as a general design technique.
SCIENTIFIC REPORTS
(2022)
Article
Computer Science, Information Systems
Akeem Pedro, Sangeun Baik, Junhyeon Jo, Doyeop Lee, Rahat Hussain, Chansik Park
Summary: This paper proposes a linked data and ontology-based system framework for effectively sharing and accessing educational safety contents from heterogeneous sources in a standardized way. Experimental results confirm that the proposed approach outperforms existing repositories in terms of search time, search result precision, and accuracy.
Article
Computer Science, Artificial Intelligence
Abhijit Adhikari, Biswanath Dutta, Animesh Dutta
Summary: Semantic Similarity research is important with high correlation using IC methods. MICA can be found in two ontologies without the need for label string matching, showing potential for new directions.
EXPERT SYSTEMS WITH APPLICATIONS
(2022)
Article
Computer Science, Information Systems
Yeonchan Ahn, Sang-Goo Lee, Junho Shim, Jaehui Park
Summary: This paper introduces a novel retrieval-augmented response generation model that retrieves relevant documents based on both the topic and local context of a conversation to generate knowledge-grounded responses. The research shows that the model can generate more knowledgeable, diverse, and relevant responses compared to existing models.
Article
Computer Science, Theory & Methods
Lynda Tamine, Lorraine Goeuriot
Summary: The explosive growth of medical information on the Internet has led to increased research activity in health informatics and information retrieval communities. Despite the low performance levels of current medical search systems, semantic search techniques offer potential to facilitate medical information retrieval. The survey also discusses key scientific challenges and potential future research directions.
ACM COMPUTING SURVEYS
(2021)
Article
Computer Science, Information Systems
Pengjun Zhai, Yu Fang, Xue Cui
Summary: With the rapid popularization of intelligent medical devices and consultation platforms, this paper proposes a question answering method based on dual-dimensional entity association for intelligent medicine to improve the accuracy and robustness of the question answering model by learning semantics from the dual-dimension of question and answer.
MOBILE INFORMATION SYSTEMS
(2022)
Article
Chemistry, Multidisciplinary
Mohamed Lemine Sidi, Serkan Gunal
Summary: This paper proposes a Purely Entity-based Semantic Search Approach for Information Retrieval (PESS4IR) to improve document retrieval. The approach includes its own entity linking and inverted indexing methods, as well as an appropriate ranking method. The experiments show that the approach achieves good performance on queries with rich annotations.
APPLIED SCIENCES-BASEL
(2023)
Article
Computer Science, Interdisciplinary Applications
Naouress Fatfouta, Julie Stal le-Cardinal
Summary: The design process in the automotive industry is expensive and time-consuming, with knowledge management playing a crucial role. This study proposes an integrated collaborative approach based on ontology for knowledge management to support simulation-aided design, specifically in car crash simulation.
COMPUTERS IN INDUSTRY
(2021)
Article
Computer Science, Information Systems
Pei-Chi Lo, Ee-Peng Lim
Summary: Contextual path retrieval (CPR) is to find contextual path(s) between a pair of entities in a knowledge graph, explaining the connection between them in a given context. We propose Embedding-based Contextual Path Retrieval (ECPR) framework, which includes a context encoder, a path encoder, and a path ranker. Our experiments on synthetic and real datasets demonstrate superior performance of ECPR-based methods over baselines, particularly with our proposed context encoders.
ACM TRANSACTIONS ON INFORMATION SYSTEMS
(2023)
Article
Computer Science, Information Systems
Yuliang Xiao, Lijuan Zhang, Jie Huang, Lei Zhang, Jian Wan
Summary: This paper introduces a knowledge graph-based question answering approach and proposes a joint system that efficiently generates and ranks candidate paths for complex questions. A new text-matching method is also introduced to capture the semantic correlation between questions and candidate paths.
Article
Green & Sustainable Science & Technology
Ho-Jin Cha, So-Won Choi, Eul-Bum Lee, Duk-Man Lee
Summary: The complexity and age of industrial plants have led to an increased need for equipment maintenance and replacement. To address the challenge of reducing the process and review time of equipment purchase order (PO) documents, a purchase order knowledge retrieval model (POKREM) was developed. POKREM utilizes knowledge graph (KG) technology and a hierarchical structure to create a graph database for accurate and efficient document search. The implementation of POKREM resulted in a significant reduction in PO document review time and improved work efficiency for engineers.
Article
Computer Science, Information Systems
Stuart J. Nelson, Allen Flynn, Mark S. Tuttle
Summary: The study aimed to develop an ontology for formalizing drug indications in a computable and comparable manner. A model was created to represent FDA-approved indications as disjuncts of conjuncts of assertions, with logical primitives chosen from 2 categories. The model successfully represented over 99% of approved treatment or prevention label indications, with challenges remaining in workflow design and terminology integration.
JOURNAL OF THE AMERICAN MEDICAL INFORMATICS ASSOCIATION
(2021)
Article
Economics
Salam Al-Mahaseneh, Yousra Harb
Summary: This study explores the factors that influence the assimilation of ICT KM practices in organizations and finds that organizational culture, structure, management support, leadership, perceived benefits, and individual attitude have significant effects on employees' assimilation of ICT KM practices.
JOURNAL OF THE KNOWLEDGE ECONOMY
(2023)
Article
Computer Science, Information Systems
Jiaxin Du, Shaohua Wang, Xinyue Ye, Diana S. Sinton, Karen Kemp
Summary: By merging existing GIS bodies of knowledge and applying deep-learning methods, researchers built a GIS knowledge graph (GIS-KG) to facilitate information retrieval. The experiments demonstrated the robust support and potential of GIS-KG in exploring emerging research themes.
INTERNATIONAL JOURNAL OF GEOGRAPHICAL INFORMATION SCIENCE
(2022)
Article
Computer Science, Artificial Intelligence
Ashish Singh Patel, Giovanni Merlino, Antonio Puliafito, Ranjana Vyas, O. P. Vyas, Muneendra Ojha, Vivek Tiwari
Summary: This paper proposes a NLP-guided approach to generate an ontology for multimedia representation and information retrieval. By leveraging textual data, possible descriptions and actions are generated, and relations among objects are embedded as object properties and classes. The completeness and coverage of the ontology are demonstrated through comparison with existing multimedia ontologies and evaluation. Spatial reasoning rules are established using SWRL rules, and information retrieval is demonstrated using DL and SPARQL queries.
EXPERT SYSTEMS WITH APPLICATIONS
(2023)
Article
Computer Science, Information Systems
Yanxiang Ling, Fei Cai, Jun Liu, Honghui Chen, Maarten de Rijke
Summary: Recent research emphasizes the importance of mixed-initiative interactions in conversational search. The task of question generation (QG) in open-domain conversational systems aims to enhance human-machine interactions. However, the limited availability of QG-specific data in conversations makes this task challenging. In this study, we propose a context-enhanced neural question generation (CNQG) model that leverages conversational context to predict question content and pattern. We also use multi-task learning with auxiliary training objectives and a self-supervised approach to train our question generator.
ACM TRANSACTIONS ON INFORMATION SYSTEMS
(2023)
Article
Economics
Olivier Sprangers, Sebastian Schelter, Maarten de Rijke
Summary: Probabilistic time series forecasting is crucial in various domains, and Transformer-based methods have achieved state-of-the-art performance. However, they require a large number of parameters and high memory requirements. To address this, we propose a novel bidirectional temporal convolutional network with significantly fewer parameters. Our method performs on par with state-of-the-art approaches and requires lower memory, reducing infrastructure cost.
INTERNATIONAL JOURNAL OF FORECASTING
(2023)
Article
Computer Science, Information Systems
Romain Deffayet, Jean-Michel Renders, Maarten De Rijke
Summary: The performance of click models under policy distributional shift (PDS) is examined, and a new evaluation protocol is proposed to predict their performance under PDS, along with guidelines to mitigate risks.
ACM TRANSACTIONS ON INFORMATION SYSTEMS
(2023)
Article
Computer Science, Information Systems
Muyang Ma, Pengjie Ren, Zhumin Chen, Zhaochun Ren, Huasheng Liang, Jun Ma, Maarten De Rijke
Summary: One of the key challenges in sequential recommendation is how to extract and represent user preferences. We propose a transformer-based sequential recommendation model, named MrTransformer, to explore multiple user preferences. MrTransformer employs preference-editing-based self-supervised learning mechanism to disentangle user preferences into multiple independent representations, improving preference extraction and representation. Experiments show that MrTransformer with preference editing outperforms state-of-the-art methods in terms of Recall, MRR, and NDCG, especially for long sequences of interactions.
ACM TRANSACTIONS ON INFORMATION SYSTEMS
(2023)
Article
Computer Science, Information Systems
Chen Wu, Ruqing Zhang, Jiafeng Guo, Maarten De Rijke, Yixing Fan, Xueqi Cheng
Summary: This article introduces the Word Substitution Ranking Attack (WSRA) task against Neural Ranking Models (NRMs), which aims to promote a target document's ranking by adding adversarial perturbations to its text. The proposed Pseudo Relevance-based ADversarial ranking Attack (PRADA) method outperforms existing attack strategies and successfully fools the NRM with small indiscernible perturbations of text.
ACM TRANSACTIONS ON INFORMATION SYSTEMS
(2023)
Article
Computer Science, Information Systems
Ming Li, Sami Jullien, Mozhdeh Ariannezhad, Maarten De Rijke
Summary: The study aims to investigate the performance of NBR methods in practical applications and proposes a new set of evaluation metrics to measure the performance of NBR models. By conducting experimental analysis on state-of-the-art NBR models, it reveals the actual progress and improvements of NBR methods in the recommendation process.
ACM TRANSACTIONS ON INFORMATION SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Wenchao Sun, Muyang Ma, Pengjie Ren, Yujie Lin, Zhumin Chen, Zhaochun Ren, Jun Ma, Maarten de Rijke
Summary: This study addresses the challenges of sequential recommendation in a context where multiple users share a single account and behavior is available in multiple domains. The proposed PSJNet network learns role-specific representations and filters out irrelevant information using a gating mechanism. It also combines split and join techniques to learn cross-domain representations. Experimental results demonstrate that PSJNet outperforms state-of-the-art baselines in terms of MRR and Recall.
IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING
(2023)
Article
Computer Science, Artificial Intelligence
Yanxiang Ling, Fei Cai, Jun Liu, Honghui Chen, Maarten de Rijke
Summary: Hierarchical context modeling is crucial for the response generation in multi-turn conversational systems. We propose a model named KS-CQ, which utilizes the Keep and Select modules to generate neighbor-aware context representation and context-enriched query representation. Extensive experiments demonstrate the effectiveness of our approach compared to state-of-the-art baselines in both automatic and human evaluations.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Fabian Ferrari, Jose van Dijck, Antal van den Bosch
Summary: In order to ensure the integrity of knowledge production, it is necessary to provide regulators and researchers with access to the training procedures of foundational models like GPT-4. Foundation models need to be open and accessible, although they are not synonymous.
NATURE MACHINE INTELLIGENCE
(2023)
Proceedings Paper
Computer Science, Information Systems
Philipp Hager, Maarten de Rijke, Onno Zoeter
Summary: Inverse-propensity scoring and neural click models are compared in this study for learning rankers from user clicks affected by position bias. Theoretical differences are explored and empirical comparisons are conducted on a prevalent evaluation setup. It is shown that both methods optimize for true document relevance when position bias is known, but small empirical differences are found when neural click models learn from shared, conflicting features.
ADVANCES IN INFORMATION RETRIEVAL, ECIR 2023, PT I
(2023)
Proceedings Paper
Computer Science, Information Systems
Mariya Hendriksen, Svitlana Vakulenko, Ernst Kuiper, Maarten de Rijke
Summary: This article investigates the reproducibility and replicability of state-of-the-art CMR results when evaluated on object-centric and scene-centric datasets. By selecting two different architectures of CMR models and evaluating them on two scene-centric datasets and three object-centric datasets, it is discovered that the reproducibility and replicability of the experimental results are problematic, and the scores obtained by the models on object-centric datasets are significantly lower than those obtained on scene-centric datasets.
ADVANCES IN INFORMATION RETRIEVAL, ECIR 2023, PT III
(2023)
Proceedings Paper
Computer Science, Information Systems
Weronika Lajewska, Krisztian Balog
Summary: This paper reports on reproducing the organizers' baseline and top participant submission at the TREC Conversational Assistance track in 2021. It highlights the challenges of reproducibility due to less strict requirements in accompanying papers. Results show key practical information is missing and indicate a smaller relative difference between baseline and top approach. The impact of pipeline components and dataset selection on system performance is explored, with findings suggesting the benefits of advanced retrieval techniques and different query rewriting methods.
ADVANCES IN INFORMATION RETRIEVAL, ECIR 2023, PT III
(2023)
Proceedings Paper
Computer Science, Information Systems
Thilina C. Rajapakse, Maarten de Rijke
Summary: Dense retrieval methods have outperformed traditional sparse retrieval methods in open-domain retrieval. However, there is a noticeable decrease in accuracy when these methods are applied to out-of-distribution and out-of-domain datasets. This may be due to the mismatch in information available to the context encoder and the query encoder during training. By training on datasets with multiple queries per passage, we show that dense passage retriever models perform better on out-of-distribution and out-of-domain test datasets compared to models trained on datasets with single query per passage.
ADVANCES IN INFORMATION RETRIEVAL, ECIR 2023, PT II
(2023)
Article
Communication
Fabian Ferrari, Jose van Dijck, Antal van den Bosch
Summary: The absence of benchmarks to examine the effectiveness of oversight mechanisms for generative AI systems is a problem for research and policy. This article introduces the conditions of industrial observability, public inspectability, and technical modifiability as structural elements for governing generative AI systems. These conditions are exemplified using the EU's AI Act, grounding the analysis of oversight mechanisms in the material properties of generative AI systems.
NEW MEDIA & SOCIETY
(2023)
Proceedings Paper
Computer Science, Artificial Intelligence
Ana Lucic, Maurits Bleeker, Sami Jullien, Samarth Bhargav, Maarten de Rijke
Summary: This paper explains the setup of a graduate-level course on Fairness, Accountability, Confidentiality, and Transparency in Artificial Intelligence (FACT-AI) at the University of Amsterdam, focusing on teaching FACT-AI concepts through reproducibility. The course involves a group project where students reproduce existing FACT-AI algorithms and write corresponding reports. The authors reflect on their experience teaching the course over two years, including during a global pandemic, and propose guidelines for teaching FACT-AI through reproducibility in graduate-level AI study programs.
THIRTY-SIXTH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE / THIRTY-FOURTH CONFERENCE ON INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE / TWELVETH SYMPOSIUM ON EDUCATIONAL ADVANCES IN ARTIFICIAL INTELLIGENCE
(2022)
