Higher-Order Situation Theory in Artificial Intelligence
Instructor
|
Roussanka Loukanova
Institute of Mathematics and Informatics, Bulgarian Academy of Sciences
Bulgaria
|
|
|
|
Brief Bio
Roussanka Loukanova is currently a Associated Researcher at Stockholm University, Sweden, and Institute of Mathematics and Informatics,
Bulgarian Academy of Sciences, Bulgaria. The focus of her research is on subjects such as: type theoretic approaches to information and information processing; type-theory of algorithms; type-theory of situated information; computerised processing of formal and natural languages; type-theoretic grammars; Constraint Based Lexicalized Grammar (CBLG) of natural language; computational syntax, semantics, and syntax-semantics interfaces in formal and natural languages. Her work targets new approaches to NLP in AI and new mathematical foundations of AI.
|
Abstract
Abstract: Situation Theory is a powerful, highly expressive, type theory of finely-grained information that is partial, underspecified, and situational. It has many applications to intelligent technologies, with most prominent application to computational semantics known as Situation Semantics. Currently, new applications of Situation Theory are proliferating, most distinctively to Artificial Intelligence, Natural Language Processing, neuroscience of language, and other areas of computational neuroscience.
The first part of the tutorial introduces the major concepts of Situation Theory, by including new developments. The second part is devoted to applications for computational semantics of natural and artificial languages in Artificial Intelligence.
Keywords: situation theory, information, situation semantics, parameters, partiality, situations, semantic types, semantic restrictions, context, agents, syntax-semantics interface.
Aims and Learning Objectives:
- To introduce the major concepts of Situation Theory;
- To provide the audience with knowledge for potential developments of Situation Theory and its applications to:
(1) Intelligent Language Processing;
(2) Various areas of Artificial Intelligence;
(3) Using Artificial Intelligence in Computational Neuroscience of Language;
(4) Using Artificial Intelligence in Computational Neuroscience;
How dissemination will be done: via ICAART, Internet announcement, email and other communications.
Target Audience: Researchers in theory of AI, Information Theory, Data Science, Language Processing, Computational Neuroscience, and other applications of AI.
Prerequisite Knowledge of Audience: There are no specific prerequisites. The tutorial provides the necessary background, concepts, and ideas. Experience with Logic and Natural Language Processing can be very resourceful.
Detailed Outline:
PART I: Situation Theory as Type-Theoretic Modelling of Information:
We introduce the major notions of Situation Theory. By examples from English language, we present abstract, theoretical concepts, without requiring background in theoretical approaches and specific areas of applications.
Parametric information is modeled by generalized, abstract objects that allow parametric components.
We introduce restricted parameters with recursive constraints that include information with situated components. Parameters can be instantiated with objects satisfying the constraints.
CONCEPTS OF SITUATION THEORY
(1) Types for situational information;
(2) Primitive relations, argument roles, saturation of arguments;
(3) Basic informational units;
(4) Situations and events;
(5) Situated propositions;
(6) Complex informational units;
(7) Complex situational types;
(8) Semantic parameters: basic, complex, restricted;
(9) Situated, partial, and parametric information;
(10) Information inference.
PART II: Application of Situation Theory to Computational Situation Semantics:
We give examples from English to represent semantics of human language. We use Situation Theory for language independent representation of semantic information. The technique can be applied to other languages and concurrently to multiple languages, in multi-language processing.
Utilization of Situation Theory allows blending linguistic with other heterogeneous information. Heterogeneous sources of information usually contribute to interpretation of language in contexts, e.g., visual and other perceptual information, agents' perspectives and attitudes, multidimensional graphical information, space-time dynamics, virtual worlds.
CONCEPTS OF SITUATION SEMANTICS:
(1) Linguistic contexts, utterances, and described situations.The situation-theoretic notion of an utterance models instances of language use, written, spoken, or by other means, e.g., gestures, graphics, virtual diagrams, etc., used by agents. Agents can be humans or technological entities, e.g., as components of equipment with integrated human language processing;
(2) Resource situations and parametric information;
(3) Abstract linguistic meanings and specific interpretations in contexts;
The informational content of language meanings is modelled as dependent on context, situations, events, and resource information. Contexts are modelled as situations with information that can be partial and parametric.
(4) Examples from English language for situation-semantics representations across syntactic categories:
- NPs: names, quantifiers, definite and indefinite NP descriptions;
- VPs with intransitive and transitive head verbs;
- Ambiguities in human language and parametric semantic content;
- Quantification scope ambiguities;
- Knowledge and attitudes.
CONCLUSIONS: Present and future of Higher-Order Situation Theory and its applications
(1) Existing and new applications to Language Processing;
(2) Potentials for applications to various areas of Artificial Intelligence;
(3) Potentials for applications to computational neuroscience of language;
(4) Potentials for applications to computational neuroscience.
Keywords
situation theory, information, situation semantics, parameters, partiality, situations, semantic types, semantic restrictions, context, agents, syntax-semantics interface
Aims and Learning Objectives
- To introduce the major concepts of Situation Theory
- To provide the audience with knowledge for potential developments of Situation Theory and its applications to:
(1) Intelligent Language Processing
(2) various areas of Artificial Intelligence.
(3) using Artificial Intelligence in Computational Neuroscience of Language
(4) using Artificial Intelligence in Computational Neuroscience
Target Audience
Researchers in theory of AI, Information Theory, Data Science, Language Processing, Computational Neuroscience, and other applications of AI
Prerequisite Knowledge of Audience
There are no specific prerequisites. The tutorial provides the necessary background, concepts, and ideas. Experience with Logic and Natural Language Processing can be very resourceful.
Detailed Outline
PART I: SITUATION THEORY AS TYPE-THEORETIC MODELLING OF INFORMATION
We introduce the major notions of Situation Theory. By examples from English language, we present abstract, theoretical concepts, without requiring background in theoretical approaches and specific areas of applications.
Parametric information is modeled by generalized, abstract objects that allow parametric components. We introduce restricted parameters with recursive constraints that include information with situated components. Parameters can be instantiated with objects satisfying the constraints.
CONCEPTS OF SITUATION THEORY
(1) Types for situational information
(2) Primitive relations, argument roles, saturation of arguments
(3) Basic informational units
(4) Situations and events
(5) Situated propositions
(6) Complex informational units
(7) Complex situational types
(8) Semantic parameters: basic, complex, restricted
(9) Situated, partial, and parametric information
(10) Information inference
PART II: Application of Situation Theory to Computational Situation Semantics.
We give examples from English to represent semantics of human language. We use Situation Theory for language independent representation of semantic information. The technique can be applied to other languages and concurrently to multiple languages, in multi-language processing.
Utilization of Situation Theory allows blending linguistic with other heterogeneous information.
Heterogeneous sources of information usually contribute to interpretation of language in contexts, e.g., visual and other perceptual information, agents' perspectives and attitudes, multidimensional graphical information, space-time dynamics, virtual worlds.
CONCEPTS OF SITUATION SEMANTICS:
(1) Linguistic contexts, utterances, and described situations
The situation-theoretic notion of an utterance models instances of language use, written, spoken, or by other means, e.g., gestures, graphics, virtual diagrams, etc., used by agents. Agents can be humans or technological entities, e.g., as components of equipment with integrated human language processing.
(2) Resource situations and parametric information
(3) Abstract linguistic meanings and specific interpretations in contexts
The informational content of language meanings is modelled as dependent on context, situations, events, and resource information. Contexts are modelled as situations with information that can be partial and parametric.
(4) Examples from English language for situation-semantics representations across syntactic categories:
- NPs: names, quantifiers, definite and indefinite NP descriptions
- VPs with intransitive and transitive head verbs
- Ambiguities in human language and parametric semantic content
- Quantification scope ambiguities
- Knowledge and attitudes
CONCLUSIONS: Present and future of Higher-Order Situation Theory and its applications
(1) Existing and new applications to Language Processing
(2) Potentials for applications to various areas of Artificial Intelligence
(3) Potentials for applications to computational neuroscience of language
(4) Potentials for applications to computational neuroscience
Secure our society - Computer Vision Techniques for Video Surveillance
Instructor
|
Huiyu Zhou
Queen's University Belfast
United Kingdom
|
|
|
|
Brief Bio
Huiyu Zhou is a Lecturer in the School of Electronics, Electrical Engineering and Computer Science at Queen’s University Belfast, United Kingdom. He obtained a Bachelor of Engineering degree in Radio Technology from the Huazhong University of Science and Technology of China, and a Master of Science degree in Biomedical Engineering from the University of Dundee of United Kingdom, respectively. He was then awarded a Doctor of Philosophy degree in Computer Vision from the Heriot-Watt University, Edinburgh, United Kingdom. He has worked in the Guangxi Medical University (China), Elscint Ltd. (Israel), University of Essex (UK), University of London (UK), and Brunel University (UK). He has taken part in the consortiums of a number of research projects in medical image processing, computer vision, intelligent systems and data mining. Dr. Zhou has published over 120 peer reviewed papers in the field. He is the recipient of CVIU 2012 Most Cited Paper Award and was shortlisted for MBEC 2006 Nightingale Prize. He serves as the Editor-in-Chief of Recent Patents on Electrical & Electronic Engineering, and is on the Editorial Boards of several refereed journals.
|
Abstract
Abstract: Research in video surveillance has received overwhelming attention in the last three decades. Increased precautions are taken in security-sensitive areas, such as country borders, airports and government offices. Individuals are also seeking personalized security systems to continuously monitor their properties and valuable assets. To meet these requirements, cameras have been deployed to acquire sensory data, followed by thorough detection and assessment of threats on-line or off-line. Understanding and interpreting object behaviours based on video analysis has witnessed impressive progress in recent years. The performance of these surveillance systems is continuously improved. In this tutorial, we will summarize recent research progress in human detection and tracking, identification, activity/behaviour analysis, motion trajectory clustering and structure from motion.
Keywords: Computer vision; machine learning; video surveillance; behaviour analysis.
Aims and Learning Objectives:
This tutorial aims to cover the early stages of the video surveillance pipeline. Participants are expected to get new perspectives within the area of video surveillance. In more detail, the objectives of this tutorial are:
I) To become acquainted with computer vision techniques used for video surveillance tasks;
II) To experience theoretical and engineering project management for video surveillance;
III) To be able to evaluate the prospective usage of video surveillance techniques for real time operation.
Target Audience: This tutorial is intended for researchers and practitioners with background in general signal/image processing.
Prerequisite Knowledge of Audience: None.
Detailed Outline: The following is a list of tentative topics and the corresponding time allocations:
a) Introduction to video surveillance
b) Object detection and tracking
c) Human profiling
d) Activity recognition
e) Trajectory clustering
f) Structure from motion