Semantically Controlled Response Generation for Task-Oriented Dialogue Systems
Prof. Maarten de Rijke, Distinguished University Professor of Artificial
Intelligence and Information Retrieval at the University of
Amsterdam, NL
Abstract
(Based on joint work with Phillip Lippe, Pengjie Ren, Hinda Haned, and Bart Voorn)
Access to information is increasingly conversational in nature. This is especially true in task-oriented settings. Response generation is a core component that translates system actions into natural language responses. In the talk I will discuss some of the challenges of balancing (i) high precision and high predictability of responses that optimize for utility and (ii) diverse and natural responses that optimize for the user experience. I will also discuss some recent advances on architectures that generate system responses by learning to paraphrase semantically constrained template-based responses.
Biography
Maarten de Rijke is a Distinguished University Professor of Artificial Intelligence and Information Retrieval at the University of Amsterdam. He is also the scientific director of the national Innovation Center for Artificial Intelligence (ICAI). His research is focused on designing trustworthy technology to connect people to information, particularly search engines, recommender systems, and conversational assistants. His work targets two key questions: How can we create intrinsic trust in information retrieval systems, that is, align their reasoning process with human expectations? And how can we establish extrinsic trust in information retrieval systems, that is, establish verifiable guarantees on their behavior?
Introduction to Enterprise knowledge graph Practice and Key Technologies
Dr. Baoxing Huai, Vice-director of Huawei Cloud Language and Speech
Innovation Lab and Chief architect of Huawei Knowledge Computing
Abstract
The development of digital transformation has driven the application of AI in the industry from perceptual intelligence to cognitive intelligence. Industry and enterprise knowledge inherent in the ocean of unstructured and semi-structured data is valuable for enterprises. The capability of mining and utilizing knowledge from those data is critical for enterprise to achieve knowledge-based transformation. By abstracting the knowledge graph construction process and bonding AI techonologies such as information extraction and KBQA algorithms, an end-to-end knowledge graph solution can be established easily to promote the application of knowledge graphs in different bussiness domains and scenarios. In this presentation, we are going to introduce the exploration and practice of knowledge computing in enterprises, including enterprise-level knowledge graph platform and knowledge extraction. In addition, we will also share the application cases of knowledge graph technology in the industry, and discuss the future technical challenges and trends of enterprise knowledge transformation.
Biography
Dr. Huai Baoxing is vice director of Huawei Cloud Language and Speech Innovation Lab and Chief architect of Huawei Knowledge Computing, graduated from University of Science and Technology of China. Dr. Huai's research interests are natural language processing and machine learning. Many of his research results have been published in AAAI/IJCAI/MM/WWW/TKDD and other top journals. Dr.Huai has a deep understanding of knowledge graph, intelligent dialogue, pre-trained language model and other artificial intelligence technologies, as well as accelerated AI landing in the industry. In addition, Dr.Huai has rich experience in algorithm development and optimization as well as the landing of business products.
Semantically Controlled Response Generation for Task-Oriented Dialogue Systems
Prof. Maarten de Rijke, Distinguished University Professor of Artificial Intelligence and Information Retrieval at the University of Amsterdam, NL
Abstract
(Based on joint work with Phillip Lippe, Pengjie Ren, Hinda Haned, and Bart Voorn)
Access to information is increasingly conversational in nature. This is especially true in task-oriented settings. Response generation is a core component that translates system actions into natural language responses. In the talk I will discuss some of the challenges of balancing (i) high precision and high predictability of responses that optimize for utility and (ii) diverse and natural responses that optimize for the user experience. I will also discuss some recent advances on architectures that generate system responses by learning to paraphrase semantically constrained template-based responses.
Biography
Maarten de Rijke is a Distinguished University Professor of Artificial Intelligence and Information Retrieval at the University of Amsterdam. He is also the scientific director of the national Innovation Center for Artificial Intelligence (ICAI). His research is focused on designing trustworthy technology to connect people to information, particularly search engines, recommender systems, and conversational assistants. His work targets two key questions: How can we create intrinsic trust in information retrieval systems, that is, align their reasoning process with human expectations? And how can we establish extrinsic trust in information retrieval systems, that is, establish verifiable guarantees on their behavior?
Introduction to Enterprise knowledge graph Practice and Key Technologies
Dr. Baoxing Huai, Vice-director of Huawei Cloud Language and Speech Innovation Lab and Chief architect of Huawei Knowledge Computing
Abstract
The development of digital transformation has driven the application of AI in the industry from perceptual intelligence to cognitive intelligence. Industry and enterprise knowledge inherent in the ocean of unstructured and semi-structured data is valuable for enterprises. The capability of mining and utilizing knowledge from those data is critical for enterprise to achieve knowledge-based transformation. By abstracting the knowledge graph construction process and bonding AI techonologies such as information extraction and KBQA algorithms, an end-to-end knowledge graph solution can be established easily to promote the application of knowledge graphs in different bussiness domains and scenarios. In this presentation, we are going to introduce the exploration and practice of knowledge computing in enterprises, including enterprise-level knowledge graph platform and knowledge extraction. In addition, we will also share the application cases of knowledge graph technology in the industry, and discuss the future technical challenges and trends of enterprise knowledge transformation.
Biography
Dr. Huai Baoxing is vice director of Huawei Cloud Language and Speech Innovation Lab and Chief architect of Huawei Knowledge Computing, graduated from University of Science and Technology of China. Dr. Huai's research interests are natural language processing and machine learning. Many of his research results have been published in AAAI/IJCAI/MM/WWW/TKDD and other top journals. Dr.Huai has a deep understanding of knowledge graph, intelligent dialogue, pre-trained language model and other artificial intelligence technologies, as well as accelerated AI landing in the industry. In addition, Dr.Huai has rich experience in algorithm development and optimization as well as the landing of business products.