Article
Engineering, Electrical & Electronic
Ryota Komatsu, Shengzhou Gao, Wenxin Hou, Mingxin Zhang, Tomohiro Tanaka, Keisuke Toyoda, Yusuke Kimura, Kent Hino, Yu Iwamoto, Kosuke Mori, Takuma Okamoto, Takahiro Shinozaki
Summary: Researchers propose spoken language acquisition agents that simulate the process of human language learning. By integrating multiple learning types, the agents successfully acquire spoken language from scratch and improve learning efficiency.
IEEE JOURNAL OF SELECTED TOPICS IN SIGNAL PROCESSING
(2022)
Article
Engineering, Industrial
Yifan Zhou, Bangcheng Li, Tian Ran Lin
Summary: This paper introduces a hierarchical coordinated reinforcement learning (HCRL) algorithm to optimize maintenance of large-scale multicomponent systems, with agent parameters and coordination relationships designed based on system characteristics, and a hierarchical structure established according to components' structural importance measures. The effectiveness of the algorithm is confirmed through validation on different systems, outperforming other methods including deep reinforcement learning.
RELIABILITY ENGINEERING & SYSTEM SAFETY
(2022)
Article
Construction & Building Technology
Ning Wang, Raja R. A. Issa, Chimay J. Anumba
Summary: This research developed a transfer learning-based text classification method using the RoBERTa neural network to accurately classify different building information-related queries into predefined categories, providing information retrieval support for a virtual assistant.
AUTOMATION IN CONSTRUCTION
(2022)
Article
Computer Science, Artificial Intelligence
Abhisek Tiwari, Sriparna Saha, Pushpak Bhattacharyya
Summary: Disease diagnosis is a crucial step in the treatment process, and automatic disease diagnosis has gained popularity due to its efficiency, accessibility, and reliability. This study proposes a knowledge-infused context-driven hierarchical reinforcement learning diagnosis dialogue system, which utilizes a Bayesian learning-inspired symptom investigation module to aid context-aware and knowledge-grounded symptom investigation. The framework also incorporates a hierarchical disease classifier to alleviate symptom state sparsity issues.
KNOWLEDGE-BASED SYSTEMS
(2022)
Article
Computer Science, Information Systems
Tulika Saha, Dhawal Gupta, Sriparna Saha, Pushpak Bhattacharyya
Summary: The paper presents a hierarchical method for efficient Dialogue Management (DM) strategy using Deep Reinforcement Learning (DRL) networks in task-oriented conversations. The system is scalable and capable of handling multiple intents, demonstrating a 41% improvement in dialogue length for a 5-intent dialogue system compared to a single-intent system.
MULTIMEDIA TOOLS AND APPLICATIONS
(2021)
Article
Automation & Control Systems
Jehyun Park, Jongeun Choi, Sungjae Nah, Dohee Kim
Summary: This paper introduces a novel approach that combines hierarchical reinforcement learning and distributional reinforcement learning to address complex sparse-reward tasks. The proposed method models random rewards as random variables following a value distribution, and uses a hierarchical policy structure. The results demonstrate the effectiveness of this method in handling uncertainties caused by noise and perturbations, and it shows potential for developing more robust and effective reinforcement learning algorithms in real physical systems.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2023)
Article
Robotics
Robert Gieselmann, Florian T. Pokorny
Summary: This study introduces a novel algorithm called PAHRL, which combines planning algorithms and reinforcement learning to address problems with implicitly defined goals by dividing tasks into shorter MDPs. During testing, a planner determines useful subgoals on the state graph constructed at the bottom level, showcasing the effectiveness of this approach in solving long-horizon decision-making problems.
IEEE ROBOTICS AND AUTOMATION LETTERS
(2021)
Article
Construction & Building Technology
Phillip Stoffel, Laura Maier, Alexander Kuempel, Thomas Schreiber, Dirk Mueller
Summary: Advanced building control strategies like model predictive control and reinforcement learning, combined with forecasts for weather, occupancy, and energy prices, have the potential to reduce buildings' energy consumption and CO2 emissions. However, there is a lack of comparability among different control algorithms in the literature. This paper extensively evaluates six advanced control algorithms based on quantitative and qualitative key performance indicators.
ENERGY AND BUILDINGS
(2023)
Article
Computer Science, Information Systems
Dimitrius F. Borges, Joao Paulo R. R. Leite, Edmilson M. Moreira, Otavio A. S. Carpinteiro
Summary: Utilizing Hierarchical Reinforcement Learning and Options Framework to control signalized intersections, this study demonstrates superior performance over fixed-time traffic controllers, offering a simple and efficient alternative for urban traffic management challenges.
Article
Chemistry, Multidisciplinary
Rong Zhou, Zhisheng Zhang, Yuan Wang
Summary: This paper proposes a hierarchical episodic control model to address the low training efficiency and high sample demand in deep reinforcement learning. By extending episodic memory to hierarchical reinforcement learning and employing a hierarchical implicit memory planning approach, the model effectively enhances training efficiency and shows notable improvements in different environments and cases of sparse rewards.
APPLIED SCIENCES-BASEL
(2023)
Review
Computer Science, Interdisciplinary Applications
Anna Procopio, Giuseppe Cesarelli, Leandro Donisi, Alessio Merola, Francesco Amato, Carlo Cosentino
Summary: Mechanistic-based models (MMs) and Machine Learning (ML) techniques are often used separately to investigate biological systems. This review investigates the combination of MMs and ML in systems biology research and highlights the great potential of this hybrid approach at both micro and macro biological scales.
COMPUTER METHODS AND PROGRAMS IN BIOMEDICINE
(2023)
Article
Computer Science, Artificial Intelligence
Sajjad Hassanpour, Vicente Gonzalez, Jiamou Liu, Yang Zou, Guillermo Cabrera-Guerrero
Summary: Post-earthquake evacuation behavior is crucial for evaluating the performance of indoor building design. However, it has been neglected in current building design practices. This paper proposes a comprehensive model that integrates human evacuation behavior and building layout design, and it has been verified through simulations.
ADVANCED ENGINEERING INFORMATICS
(2022)
Article
Computer Science, Theory & Methods
Shubham Pateria, Budhitama Subagdja, Ah-hwee Tan, Chai Quek
Summary: Hierarchical Reinforcement Learning (HRL) allows for autonomous decomposition of challenging decision-making tasks, with a growing landscape of research approaches. Emphasis is placed on learning strategies, subtask discovery, transfer learning, and multi-agent learning challenges, with proposed open problems and practical application examples highlighted.
ACM COMPUTING SURVEYS
(2021)
Article
Engineering, Industrial
Junliang Wang, Pengjie Gao, Peng Zheng, Jie Zhang, W. H. Ip
Summary: The paper introduces a fuzzy hierarchical reinforcement learning approach for scheduling semiconductor wafer manufacturing systems to improve on-time delivery. By utilizing a hierarchical model and recurrent reinforcement learning units to address layer and wafer correlation, the control of cycle time is achieved.
JOURNAL OF MANUFACTURING SYSTEMS
(2021)
Article
Mechanics
S. Rezaeiravesh, T. Mukha, P. Schlatter
Summary: Multifidelity models (MFMs) are used to construct predictive models for flow quantities of interest (QoIs) over uncertain/design parameters, for the purpose of uncertainty quantification, data fusion, and optimization. The hierarchical MFM strategy allows for simultaneous calibration of fidelity-specific parameters in a Bayesian framework, combining lower and higher-fidelity data in an optimal way to provide improved prediction and confidence intervals for QoIs.
JOURNAL OF FLUID MECHANICS
(2023)
Article
Computer Science, Artificial Intelligence
Heriberto Cuayahuitl
Article
Computer Science, Artificial Intelligence
Oliver Lemon
Summary: Research at the Interaction Lab focuses on human-agent communication using conversational Natural Language. The goal is to create systems where humans and AI agents can form teams and coordinate tasks through Natural Language conversation. This paper introduces machine learning approaches to conversational AI and covers practical systems developed in the lab, including communication between multiple agents. It also discusses future directions for conversational, collaborative multi-agent systems.
Article
Acoustics
Erfan Loweimi, Zhengjun Yue, Peter Bell, Steve Renals, Zoran Cvetkovic
Summary: In this paper, the authors investigate multi-stream acoustic modelling using the raw real and imaginary parts of the Fourier transform of speech signals. They discuss the importance of such information and propose a framework where the real and imaginary parts are treated as separate streams and combined at an optimal level of abstraction. The proposed systems achieved competitive performance in various tasks, including phone recognition, noise robustness, and speech intelligibility.
IEEE-ACM TRANSACTIONS ON AUDIO SPEECH AND LANGUAGE PROCESSING
(2023)
Proceedings Paper
Engineering, Electrical & Electronic
Meriam Moujahid, Helen Hastie, Oliver Lemon
Summary: This study introduces a multi-user engagement strategy that utilizes the robot's gaze, head pose, and verbal communication to coordinate turn-taking and analyzes the participants' perception of the robot. The results confirm that the robot is perceived as more intelligent and conscious when it reacts using eye gaze or head pose when a new user enters the scene. Furthermore, it is found that robots need to use a combination of verbal and non-verbal cues to coordinate turn-taking in order to be perceived as polite and aware of human social norms.
PROCEEDINGS OF THE 2022 17TH ACM/IEEE INTERNATIONAL CONFERENCE ON HUMAN-ROBOT INTERACTION (HRI '22)
(2022)
Proceedings Paper
Engineering, Electrical & Electronic
Meriam Moujahid, Bruce Wilson, Helen Hastie, Oliver Lemon
Summary: The demonstration showcases a Robot Receptionist that can handle multi-party engagement and turn-taking in dynamic environments. Utilizing a highly expressive Furhat robot, the system consists of several modules including scene analysis, engagement policies, and a dialogue manager.
PROCEEDINGS OF THE 2022 17TH ACM/IEEE INTERNATIONAL CONFERENCE ON HUMAN-ROBOT INTERACTION (HRI '22)
(2022)
Proceedings Paper
Computer Science, Cybernetics
Nancie Gunson, Daniel Hernandez Garcia, Weronika Sieinska, Christian Dondrup, Oliver Lemon
Summary: This paper describes the potential applications of social robots in healthcare settings, such as robot receptionists, to assist patients and visitors and alleviate staff workload. It presents the development of a multimodal conversational AI system integrated in a social conversational robot (ARI robot) and reports on an initial experimental validation study conducted with the ARI robot in laboratory conditions.
2022 31ST IEEE INTERNATIONAL CONFERENCE ON ROBOT AND HUMAN INTERACTIVE COMMUNICATION (IEEE RO-MAN 2022)
(2022)
Article
Acoustics
Dino Oglic, Zoran Cvetkovic, Peter Sollich, Steve Renals, Bin Yu
Summary: This study focuses on the problem of learning robust acoustic models in adverse environments. The authors propose using data augmentation as a way to improve risk estimates during training and demonstrate its effectiveness through theoretical analysis and empirical results.
IEEE-ACM TRANSACTIONS ON AUDIO SPEECH AND LANGUAGE PROCESSING
(2022)
Proceedings Paper
Computer Science, Artificial Intelligence
Alessandro Suglia, Yonatan Bisk, Ioannis Konstas, Antonio Vergari, Emanuele Bastianelli, Andrea Vanzo, Oliver Lemon
Summary: Guessing games serve as a prototypical example of the learning by interacting paradigm, and this research investigates how artificial agents can benefit from playing such games in the context of NLP tasks. The study proposes two methods, supervised learning and self-play via SPIEL, to exploit guessing games, and evaluates their generalization ability to improve performance in downstream NLP tasks. The results show increased accuracy in both in-domain and transfer evaluations, with SPIEL leading to more fine-grained object representations for improved performance in VQA.
16TH CONFERENCE OF THE EUROPEAN CHAPTER OF THE ASSOCIATION FOR COMPUTATIONAL LINGUISTICS (EACL 2021)
(2021)
Proceedings Paper
Computer Science, Artificial Intelligence
Jose L. Part, Daniel Hernandez Garcia, Yanchao Yu, Nancie Gunson, Christian Dondrup, Oliver Lemon
Summary: The goal of the SPRING project is to develop a socially pertinent robot for tasks in gerontological healthcare. The robot must be able to perceive its environment and have coherent conversations about the surrounding world. The described work has applications beyond healthcare and can be used on any robot that needs to interact with its visual and spatial environment.
HRI '21: COMPANION OF THE 2021 ACM/IEEE INTERNATIONAL CONFERENCE ON HUMAN-ROBOT INTERACTION
(2021)
Proceedings Paper
Computer Science, Artificial Intelligence
Nancie Gunson, Weronika Sieinska, Christopher Walsh, Christian Dondrup, Oliver Lemon
PROCEEDINGS OF THE 20TH ACM INTERNATIONAL CONFERENCE ON INTELLIGENT VIRTUAL AGENTS (ACM IVA 2020)
(2020)
Proceedings Paper
Computer Science, Artificial Intelligence
Weronika Sieinska, Nancie Gunson, Christopher Walsh, Christian Dondrup, Oliver Lemon
SIGDIAL 2020: 21ST ANNUAL MEETING OF THE SPECIAL INTEREST GROUP ON DISCOURSE AND DIALOGUE (SIGDIAL 2020)
(2020)
Proceedings Paper
Computer Science, Interdisciplinary Applications
Andrea Vanzo, Emanuele Bastianelli, Oliver Lemon
20TH ANNUAL MEETING OF THE SPECIAL INTEREST GROUP ON DISCOURSE AND DIALOGUE (SIGDIAL 2019)
(2019)
Proceedings Paper
Computer Science, Interdisciplinary Applications
Igor Shalyminov, Sungjin Lee, Arash Eshghi, Oliver Lemon
20TH ANNUAL MEETING OF THE SPECIAL INTEREST GROUP ON DISCOURSE AND DIALOGUE (SIGDIAL 2019)
(2019)
Proceedings Paper
Computer Science, Artificial Intelligence
Jose L. Part, Oliver Lemon
2019 IEEE/RSJ INTERNATIONAL CONFERENCE ON INTELLIGENT ROBOTS AND SYSTEMS (IROS)
(2019)
Proceedings Paper
Acoustics
Shucong Zhang, Erfan Loweimi, Peter Bell, Steve Renals
2019 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH AND SIGNAL PROCESSING (ICASSP)
(2019)
Article
Computer Science, Artificial Intelligence
Yanyue Zhang, Deyu Zhou, Zhenglin Wang, Yilong Lai
Summary: This paper proposes an unsupervised opinion summarization method that addresses the problem of generating inaccurate content through adversarial learning, without requiring specific model structures or domain metadata. By appending natural language inference as the discriminator to the generation model and retraining the discriminator for unsupervised contrastive learning, the model achieves model-agnostic and metadata-free performance. Experimental results demonstrate that the proposed method generates comparable results on ROUGE scores and outperforms state-of-the-art baselines in category accuracy and sentiment accuracy for summarization faithfulness evaluation.
COMPUTER SPEECH AND LANGUAGE
(2024)
Article
Computer Science, Artificial Intelligence
Loredana Schettino, Antonio Origlia, Francesco Cutugno
Summary: This study presents the results of two perception experiments that evaluate the impact of specific patterns of disfluencies on listeners of synthetic speech. Focusing on Cultural Heritage presentations, the study proposes a linguistic model for positioning disfluencies in Italian language utterances. Utilizing a state-of-the-art speech synthesizer based on Deep Neural Networks, the study prepares experimental stimuli and conducts subjective evaluations and behavioral assessments. The results indicate that synthetic utterances with predicted disfluencies are perceived as more natural and improve the listeners' recall of the provided information.
COMPUTER SPEECH AND LANGUAGE
(2024)
Article
Computer Science, Artificial Intelligence
Ying Zhang, Wencheng Huang, Depeng Dang
Summary: This paper introduces a lightweight approach to address the problem of few-shot relation extraction, using prompt-learning to assist in fine-tuning the model and designing an enhanced fusion module to fuse relation information and original prototype. Experimental results show that the proposed method achieves state-of-the-art performance on common datasets.
COMPUTER SPEECH AND LANGUAGE
(2024)
Article
Computer Science, Artificial Intelligence
Sebastien Le Maguer, Simon King, Naomi Harte
Summary: The release of WaveNet and Tacotron has greatly impacted the speech synthesis field by significantly improving the quality of synthetic speech. However, the current evaluation protocol, Absolute Category Rating (ACR) and Mean Opinion Score (MOS), used to measure this quality, has sparked controversy. To determine the reliability of MOS, a series of experiments were conducted, examining the stability of MOS over time, the influence of lower quality systems on MOS, the influence of modern technologies on past system scores, and the evolution of MOS for modern technologies in isolation. The results suggest the need for new evaluation protocols better suited for analyzing modern speech synthesis technologies.
COMPUTER SPEECH AND LANGUAGE
(2024)
Article
Computer Science, Artificial Intelligence
Yuxian Wan, Wenlin Zhang, Zhen Li, Hao Zhang, Yanxia Li
Summary: In this paper, a new knowledge distillation method called Dual Knowledge Distillation (DKD) is proposed to better utilize monolingual and limited bilingual data. By combining self-distillation and consistency regularization strategies, significant improvements are achieved in extracting consistent monolingual representation and forcing the decoder to produce consistent output.
COMPUTER SPEECH AND LANGUAGE
(2024)
Article
Computer Science, Artificial Intelligence
Kamini Sabu, Preeti Rao
Summary: Reading is a foundational skill that is given great importance in education systems across countries. The assessment of linguistic competence through oral reading has been the focus of scientific studies, connecting the reader's comprehension to various measures of oral reading fluency. As this assessment requires significant time and resources, there is interest in automating the prediction of reading fluency using the same pedagogical rubrics. This study discusses new approaches to modeling prosody for automatic assessment, highlighting the importance of prosodic features informed by speech rate and speaking style in system performance.
COMPUTER SPEECH AND LANGUAGE
(2024)
Article
Computer Science, Artificial Intelligence
Imran Sheikh, Emmanuel Vincent, Irina Illina
Summary: This article studies the training and adaptation of recurrent neural network (RNN) language models (LM) on a limited amount of in-domain speech data. It proposes training loss methods based on Kullback-Leibler (KL) divergence, hidden Markov model (HMM), and sampled paths from ASR confusion networks. Experimental results on telephone and meeting conversation datasets show that the sampling method for training RNN LMs on ASR confusion networks performs the best and leads to a relative reduction in perplexity.
COMPUTER SPEECH AND LANGUAGE
(2024)