Article
Computer Science, Hardware & Architecture
Cesar Gonzalez-Mora, Cristina Barros, Irene Garrig, Jose Zubcoff, Elena Lloret, Jose -Norberto Maz
Summary: This paper proposes a novel approach to automatically generate interactive Web API documentation. By applying natural language processing techniques, the API documentation is transformed into easily understandable natural language descriptions. Through a web interface, the documentation becomes interactive, enhancing the usability and reusability of the Web API.
COMPUTER STANDARDS & INTERFACES
(2023)
Article
Biotechnology & Applied Microbiology
Stephen Goldrick, Haneen Alosert, Clare Lovelady, Nicholas J. Bond, Tarik Senussi, Diane Hatton, John Klein, Matthew Cheeks, Richard Turner, James Savery, Suzanne S. Farid
Summary: Cell line development is a crucial stage in biopharmaceutical development, and failure to fully characterize the lead clone during initial screening can lead to delays and compromise manufacturing success. This study proposes a novel cell line development methodology called CLD (4), which uses four steps to autonomously select the lead clone based on data. CLD (4) incorporates digitalization, machine learning, and natural language generation to generate an automated report summarizing relevant statistics.
FRONTIERS IN BIOENGINEERING AND BIOTECHNOLOGY
(2023)
Article
Computer Science, Interdisciplinary Applications
Vito Giordano, Elena Coli, Antonella Martini
Summary: This paper aims to demonstrate that text mining can help make a complex open database more effective for the engineering design process, using the U.S. Open Government Data repository as a case study. It explores the expertise and data science methods required for processing different data formats, and presents significant implications and challenges for researchers, practitioners, and policy makers in the fields of engineering design and data science.
COMPUTERS IN INDUSTRY
(2022)
Article
Computer Science, Information Systems
Yang Zhou, Chenjiao Zhi, Feng Xu, Weiwei Cui, Huaqiong Wang, Aihong Qin, Xiaodiao Chen, Yaqi Wang, Xingru Huang
Summary: This paper proposes a method based on Keyword-Aware Transformers Network (KAT) to fuse contextual keywords, enabling keyword semantic enhancement. Experimental results on two Chinese open-domain dialogue datasets show that our model outperforms existing methods in both semantic and non-semantic evaluation metrics, improving Coherence, Fluency, and Informativeness in manual evaluation.
Article
Chemistry, Multidisciplinary
Heng Gong, Xiaocheng Feng, Bing Qin
Summary: Data-to-text generation is important in natural language processing for generating user-friendly descriptive text from structured data. Distantly-supervised data-to-text generation has been proposed to overcome the lack of annotated training corpus, but it suffers from over-generation due to the inclusion of hallucinated expressions. We address this issue by empowering the neural data-to-text model with meta learning to assign higher weights to well-aligned instances and rewrite low-quality texts. Experiments show that our model outperforms the state-of-the-art model in terms of automatic evaluation metrics and human evaluation. We also introduce a new dataset, DIST-ToTTo, for distantly-supervised data-to-text generation.
APPLIED SCIENCES-BASEL
(2023)
Article
Engineering, Multidisciplinary
Yangyang Zhang, Weijun Xu, Xia Zhang, Liping Li
Summary: This paper proposes a sign language recognition system that combines a data glove and a camera to detect and recognize gestures using deep learning. The system is able to recognize letters in sign language in a short amount of time, aiding in the understanding of gestures as words and texts.
Article
Computer Science, Artificial Intelligence
Elham Seifossadat, Hossein Sameti
Summary: This paper proposes a sequence-to-sequence model called DM-NLG for data-to-text generation, which can generate natural language text from structured nonlinguistic input. By adding a dynamic memory module to the attention-based sequence-to-sequence model, it can store the information leading to the generation of previous output words and use it for generating the next word, thus preventing the generation of duplicate words or incomplete semantic concepts. Experiments on five different datasets show that the proposed DM-NLG model can reduce the slot error rate by 50% and improve BLEU by 10% compared to state-of-the-art models.
NATURAL LANGUAGE ENGINEERING
(2023)
Article
Computer Science, Information Systems
Heng Gong, Xiaocheng Feng, Bing Qin
Summary: Data-to-text generation is an important task in natural language processing, aiming to provide user-friendly text to help people understand structured data. Existing methods have shown promising results in addressing content planning and surface realization challenges. However, they lack an iterative refinement process for text generation. This paper explores enhancing data-to-text generation with an iterative refinement process via diffusion.
Article
Computer Science, Artificial Intelligence
Nina Dethlefs, Annika Schoene, Heriberto Cuayahuitl
Summary: This article proposes a novel divide-and-conquer approach to automatically induce a hierarchy of generation spaces for concept-to-text generation, which performs well in experiments and overcomes issues of data sparsity. The approach outperforms flat baselines and previous work by up to 30%.
Review
Computer Science, Artificial Intelligence
Silvia Casola, Alberto Lavelli
Summary: This paper surveys Natural Language Processing (NLP) approaches in summarizing, simplifying, and generating patent text. It highlights the challenges posed by the unique characteristics of patents to the current state of NLP research, presents previous work and its evolution critically, and draws attention to areas where further research is needed. To the best of the authors' knowledge, this is the first survey of generative approaches in the patent domain.
EXPERT SYSTEMS WITH APPLICATIONS
(2022)
Article
Computer Science, Theory & Methods
Amit Kumar Jaiswal, Prayag Tiwari, Sahil Garg, M. Shamim Hossain
Summary: Researchers propose a deep learning approach based on capsule networks for named entity recognition, which can accurately predict entities in a corpus and serve as input for subsequent NLP tasks.
FUTURE GENERATION COMPUTER SYSTEMS-THE INTERNATIONAL JOURNAL OF ESCIENCE
(2021)
Article
Computer Science, Artificial Intelligence
Jianyu Zhao, Zhiqiang Zhan, Tong Li, Rang Li, Changjian Hu, Siyun Wang, Yang Zhang
Summary: Table-to-Text generation models suffer from issues like nonfluency and divergence. To address these problems, a novel GAN-based model is proposed, which trains a generative model and a discriminative model simultaneously. The model achieves significant improvements in fluency and information consistency, outperforming baselines on various evaluation metrics while also constructing a new dataset to advance research in the field.
Article
Computer Science, Artificial Intelligence
Elham Seifossadat, Hossein Sameti
Summary: Data-to-Text Generation (D2T) is a sub-field of Natural Language Generation that transcribes structured data into natural language text. This work proposes a stochastic corpus-based model that generates tree-structured sentences based on dependency information, resulting in fluent sentences with correct grammatical structures. The model improves BLEU scores and generates high-quality utterances in various domains.
COMPUTER SPEECH AND LANGUAGE
(2022)
Article
Computer Science, Theory & Methods
Wenhao Yu, Chenguang Zhu, Zaitang Li, Zhiting Hu, Qingyun Wang, Heng Ji, Meng Jiang
Summary: The goal of text-to-text generation is to make machines express like a human in various applications. However, current text generation models often lack sufficient knowledge to generate desirable outputs. To address this issue, researchers have explored integrating internal and external knowledge into text generation systems. This survey provides a comprehensive review of the research on knowledge-enhanced text generation, including general methods and architectures, as well as specific techniques and applications.
ACM COMPUTING SURVEYS
(2022)
Article
Computer Science, Artificial Intelligence
Robert Dale
Summary: Since the release of ChatGPT at the end of November 2022, there has been extensive discussion about generative AI in both the technical and mainstream media. The impact of large language model technology has been debated, from disrupting search engines to affecting student essays and spreading disinformation. This article examines how major players in the field are reacting and explores potential future developments.
NATURAL LANGUAGE ENGINEERING
(2023)
Article
Computer Science, Artificial Intelligence
Ahmet Soylu, Oscar Corcho, Brian Elvesaeter, Carlos Badenes-Olmedo, Tom Blount, Francisco Yedro Martinez, Matej Kovacic, Matej Posinkovic, Ian Makgill, Chris Taggart, Elena Simperl, Till C. Lech, Dumitru Roman
Summary: Public procurement is a significant market that impacts various organizations and individuals, requiring governments to ensure transparency, efficiency, and healthy competition. Initiatives like open data and cross-border data integration have the potential to enhance competition and lower barriers for smaller suppliers, but existing challenges include technical heterogeneity, data quality issues, and insufficient metadata.
Article
Computer Science, Artificial Intelligence
Marlene Goncalves, David Chaves-Fraga, Oscar Corcho
Summary: This paper presents a framework called Morph-Skyline++ for processing SPARQL qualitative preference queries by directly querying relational databases. The framework achieves high performance and accurately identifies the result set.
Article
Geography
Victor Saquicela, Luis M. Vilches-Blazquez, Renan Freire, Oscar Corcho
Summary: This article presents an approach for automatically generating semantic annotations of Web Feature Services (WFS) at different request levels to generate knowledge graphs, using external services, ontological resources, and knowledge bases. The approach allows for validating the annotations and demonstrates its feasibility through an application case.
TRANSACTIONS IN GIS
(2022)
Article
Computer Science, Information Systems
Ahmet Soylu, Oscar Corcho, Brian Elvesaeter, Carlos Badenes-Olmedo, Francisco Yedro-Martinez, Matej Kovacic, Matej Posinkovic, Mitja Medvescek, Ian Makgill, Chris Taggart, Elena Simperl, Till C. Lech, Dumitru Roman
Summary: Open data and knowledge graph can be powerful tools for governments to prevent fraud and corruption, improve transparency, and enhance policy making. This article presents a case study in Slovenia where anomaly detection techniques were successfully applied to detect fraud and uncompetitive markets using integrated open data sets and a linked data-based platform called TheyBuyForYou. The article also provides guidelines for publishing high quality procurement data and emphasizes the importance of data quality in procurement analytics.
Article
Computer Science, Artificial Intelligence
Alberto Barbado, Oscar Corcho, Richard Benjamins
Summary: This paper focuses on the black box problem in unsupervised learning, evaluating rule extraction techniques from OneClass SVM models and proposing algorithms for computing XAI-related metrics. The research evaluates the proposals with different data sets, including real-world data, aiming to extend XAI techniques to unsupervised machine learning models.
EXPERT SYSTEMS WITH APPLICATIONS
(2022)
Article
Computer Science, Artificial Intelligence
Ahmad Alobaid, Oscar Corcho
Summary: This paper presents a novel approach to automatically assigning ontology classes to entity columns in tabular data, without the need for external linguistic resources, lookup services, model training, building a model of the knowledge graph beforehand, or human involvement.
KNOWLEDGE-BASED SYSTEMS
(2022)
Article
Chemistry, Analytical
Nourdine Aliane, Carlos Quiterio Gomez Munoz, Javier Sanchez-Soriano
Summary: This paper proposes the development of a Web and MATLAB-based application that integrates several services in the same environment, solving the issues of task development and software integration in precision agriculture.
Article
Automation & Control Systems
Alberto Barbado, Oscar Corcho
Summary: This study combines unsupervised anomaly detection techniques, domain knowledge, and interpretable machine learning models to explain abnormal fuel consumption in vehicle fleets. Results evaluated on real-world data show that this approach provides recommendations for fuel optimization adjusted to different user profiles.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2022)
Article
Environmental Sciences
Javier Rodriguez-Vazquez, Miguel Fernandez-Cortizas, David Perez-Saura, Martin Molina, Pascual Campoy
Summary: This paper proposes a novel semi-supervised approach for accurate counting and localization of tropical plants in aerial images without labeled data. The approach utilizes deep learning and domain adaptation to handle the domain shifts between training and test data, which is a common challenge in agricultural applications. By using unsupervised domain alignment and pseudolabeling, the method adapts a model trained on a labeled source dataset to an unlabeled target dataset. Experimental results demonstrate the effectiveness of this approach in counting pineapple plants in aerial images under significant domain shifts, achieving a reduction in counting error of up to 97% (1.42 in absolute count) compared to the supervised baseline (48.6 in absolute count).
Article
Remote Sensing
Jorge Cujo Blasco, Sergio Bemposta Rosende, Javier Sanchez-Soriano
Summary: This study presents a real-time 3D reconstruction system using drones, which leverages innovative AI techniques to achieve accurate and efficient reconstruction of 3D environments. By integrating vision, navigation, and 3D reconstruction subsystems, the system overcomes the limitations of existing applications and software in terms of speed and accuracy. The proposed system outperforms traditional software by more than 90 times and contributes to the advancement in the field of 3D reconstruction using drones.
Article
Computer Science, Information Systems
Sergio Bemposta Rosende, Javier Fernandez-Andres, Javier Sanchez-Soriano
Summary: This study proposes a technique to reduce CNN training time by using an algorithm to partition the dataset and discard unnecessary objects. The average reduction in training time is 75% without altering the training results. This tool is particularly effective for sequential images, large images, and images with small targets.
Proceedings Paper
Computer Science, Artificial Intelligence
Paola Espinoza-Arias, Daniel Garijo, Oscar Corcho
Summary: This paper introduces a method for API generation based on ontologies. The authors also developed a tool to support the process and address limitations. Future work is needed to further exploit the potential of KGs and ontologies.
KNOWLEDGE ENGINEERING AND KNOWLEDGE MANAGEMENT, EKAW 2022
(2022)
Article
Remote Sensing
Mohammad Sadeq Ale Isaac, Marco Andres Luna, Ahmed Refaat Ragab, Mohammad Mehdi Ale Eshagh Khoeini, Rupal Kalra, Pascual Campoy, Pablo Flores Pena, Martin Molina
Summary: This paper introduces a medium-scale hexacopter, called the Fan Hopper, which investigates the optimum control possibilities for a fully autonomous mission carrying a heavy payload. The research reveals that tuned Electric Ducted Fan (EDF) engines function dramatically for large payloads.
Article
Computer Science, Information Systems
Enrique Puertas, Gonzalo De-Las-Heras, Javier Sanchez-Soriano, Javier Fernandez-Andres
Summary: This dataset consists of Spanish road images with annotations of Variable Message Signals, which can be used for training computer vision algorithms. It contains 1216 instances for research in road computer vision.
Article
Computer Science, Information Systems
Enrique Puertas, Gonzalo De-Las-Heras, Javier Fernandez-Andres, Javier Sanchez-Soriano
Summary: This article introduces a dataset of Spanish roundabouts aerial images with vehicle position annotations. The dataset contains 985,260 instances and can be used for training computer vision models, such as convolutional neural networks.