Marc Cavazza | Teesside University (original) (raw)
Papers by Marc Cavazza
La compréhension automatique du langage naturel nécessite un grand nombre de connaissances. Des c... more La compréhension automatique du langage naturel nécessite un grand nombre de connaissances. Des connaissances linguistiques sont indispensables pour construire une représentation informatique de la phase et des connaissances sur le monde de référence permettent d'effectuer les inférences indispensables à une bonne compréhension. Nous décrivons la réalisation d'un analyseur sémantique qui aborde ces deux aspects de la compréhension automatique en s' inspirant d'une théorie linguistique, la sémantique ...
Digital Convergence: The Information Revolution, 1999
U of Teesside, Feb 1, 2002
The simultaneous lightness contrast (SLC) is the condition whereby a grey patch on a dark backgro... more The simultaneous lightness contrast (SLC) is the condition whereby a grey patch on a dark background appears lighter than an equal patch on a light background. Interestingly, the lightness difference between the grey patches increases when the two backgrounds—although kept at the same luminance average—are patterned shaping what can be named the articulated-SLC. There are two main interpretations for these phenomena. The framework approach maintains that the visual system groups the luminance within a set of ...
Intellectica, 1996
Résumé/Abstract Les rapports entre le contenu sémantique et les formalismes de description sont u... more Résumé/Abstract Les rapports entre le contenu sémantique et les formalismes de description sont un problème majeur pour le développement d'une technologie informatique des objets culturels, dont le traitement automatique des textes n'est qu'une première étape. Dans cet article, en prenant comme cadre théorique la sémiotique du texte, nous passons en revue quelques problèmes classiques liés à la description du contenu sémantique. Ceci nous amène à avancer un certain nombre de directions de recherche pour aborder les ...
Seventh Intelligent Narrative Technologies Workshop, Oct 23, 2014
Virtual Environments (VE) can be used to represent complex situations. It is thus a logical evolu... more Virtual Environments (VE) can be used to represent complex situations. It is thus a logical evolution to extend them to carry out problem solving tasks as well. Intelligent Virtual Environments (IVE) integrate AI techniques with real-time 3D environments. As such, they can support interactive problem solving, provided the underlying AI techniques can produce solutions within a time frame matching that of the user's interaction. In these systems, the visual space is directly interfaced to a problem solver. Consequently, interacting with objects in the VE serves as input to the problem solver, which outputs new solutions directly as object configurations. This kind of system has many potential applications in design, configuration, situation assessment, etc. We have implemented an intelligent virtual environment based on constraint logic programming (GNU Prolog), integrated in a realtime 3D graphic environment ( Unreal Tournament TM ) using an event-based approach. More specifically, physical interaction with the virtual world objects can be translated into real-time input to the CLP solver by selecting the variables whose values have been altered by the interaction. For instance, when visualising a configuration, the user can alter the position of certain objects. This modifies the constraints involving these objects, triggering the solver on a new set of variables. Seen from the user's perspective, the 3D environment automatically reconfigures itself by reacting to his/her interactions. This form of interactive problem solving is illustrated in the figure below. We demonstrate the system behaviour on a configuration example, which uses real-world design knowledge in terms of building interior design for offices (a bank agency in our case). In this system, the domain knowledge is expressed in the form of constraints in the object's attributes, which correspond not only to geometric constraints but also semantic constraints (such as compatibility between materials, lightning, etc.) .The system will propose a first result configuration, which can be interactively refined by the user, through direct physical modification in the virtual world, triggering the computation of a new solution compatible with the user's input. Our results show that the sampling rate of object manipulation in the virtual environment is compatible with the result production of the solver (on average, the total time is less than 0.15 seconds), hence supporting interactive exploration of the solution space.
This paper presents the first prototype of a reconfigurable Virtual Environment (VE). The objecti... more This paper presents the first prototype of a reconfigurable Virtual Environment (VE). The objective of the system is to link 3D Intelligent Virtual Environments to interactive planning systems. This type of system makes possible interactive solutions where the user refines a possible configuration and enables the system to generate a complete new solution enforcing all the design constraints, previously programmed. In this first prototype we link our constraint solver with the visualization engine so that the solution produced by the constraint solver is displayed in a VE.
In this paper, we discuss technical issues in the implementation of Intelligent Virtual Environme... more In this paper, we discuss technical issues in the implementation of Intelligent Virtual Environments (IVE), taking as a starting point their requirement for a specific Knowledge Representation layer. As IVE should combine both an abstract level of representation and a concrete model of some physical world, with which the user can directly interact, there is a need for unifying principles that can describe the relations between the user, the Knowledge Representation layer and the Virtual World. We suggest that, while the user can still interact physically with the Virtual Environment, he can also exchange abstract information with the underlying Knowledge Representation layer through Natural Language, and that this information should be interpreted into corresponding state of affairs in the Virtual Environment. Conversely, there is also a need for describing the Virtual World dynamics in terms of the high-level concepts that populate the Knowledge Representation layer. After illustrating these issues in the context of "Brightown" scenarios, we discuss the use of some Artificial Intelligence techniques and formalisms in the specific context of Intelligent Virtual Environments and outline a framework for their implementation.
A fundamental issue when building a semantic lexicon for a Natural Language Understanding (NLU) s... more A fundamental issue when building a semantic lexicon for a Natural Language Understanding (NLU) system hinges on the amount of world knowledge that should be incorporated into the lexicon: the so-called dictionary-encyclopedia controversy. This issue can nd a technical solution if the NLU architecture separates a linguistic step and a domain inference step. In that case, the lexicon can contain purely linguistic elements. This requires a good descriptive linguistic theory, and an explicit link between knowledge representation and linguistic data. We hereby report preliminary results from a program embedding these principles.
In this paper, we report work in progress in the development of a novel platform for Virtual Real... more In this paper, we report work in progress in the development of a novel platform for Virtual Reality Art installations. We first introduce the notion of Alternative Reality that extends traditional VR to support the more radical experiments required by artistic productions. We subsequently introduce several technical approaches through which traditional laws of physics or more fundamental laws such as causality can be redefined within a virtual environment. We use a game engine as a development environment, which supports the redefinition of low-level physical behaviours. Changing physical or causal laws however requires the development of an explicit representation of high-level behaviour. Working with causality in particular calls for the use of representations developed for instance in qualitative physics. We illustrate the work in progress with early results from small-scale examples and finally describe the target platform onto which final installations will be developed.
Studies in Computational Intelligence, 2007
Summary. Cardiac emergency is one field in medicine widely studied and very often experienced in ... more Summary. Cardiac emergency is one field in medicine widely studied and very often experienced in reality. Furthermore, simulation and training issues on this field require sensitive and expensive techniques. In the following, we describe the development of a virtual human to be used for training applications in the field of cardiac emergencies. The system integrates Artificial Intelligence (AI) techniques for simulating medical conditions (shock states) with a realistic visual simulation of the patient in a 3D environment ...
Lecture Notes in Computer Science, 1998
Avatars and Artificial Actors in Virtual Environments can be controlled by speech, as an alternat... more Avatars and Artificial Actors in Virtual Environments can be controlled by speech, as an alternative to motion capture techniques. In this paper, we discuss some specific requirements for the successful implementation of speech-based control of guided actors. We describe our sublanguage approach to speech-based control and its associated parsing techniques, based on lexicalised grammars. After an introduction to the REALISM animation software, we report work in progress in the real-time processing of spoken commands ...
International Conference on Multimodal Interfaces and the Workshop on Machine Learning for Multimodal Interaction on - ICMI-MLMI '10, 2010
In this paper, we investigate the user's eye gaze behavior during the conversation with an intera... more In this paper, we investigate the user's eye gaze behavior during the conversation with an interactive storytelling application. We present an interactive eye gaze model for embodied conversational agents in order to improve the experience of users participating in Interactive Storytelling. The underlying narrative in which the approach was tested is based on a classical XIX th century psychological novel: Madame Bovary, by Flaubert. At various stages of the narrative, the user can address the main character or respond to her using free-style spoken natural language input, impersonating her lover. An eye tracker was connected to enable the interactive gaze model to respond to user's current gaze (i.e. looking into the virtual character's eyes or not). We conducted a study with 19 students where we compared our interactive eye gaze model with a non-interactive eye gaze model that was informed by studies of human gaze behaviors, but had no information on where the user was looking. The interactive model achieved a higher score for user ratings than the non-interactive model. In addition we analyzed the users' gaze behavior during the conversation with the virtual character.
La compréhension automatique du langage naturel nécessite un grand nombre de connaissances. Des c... more La compréhension automatique du langage naturel nécessite un grand nombre de connaissances. Des connaissances linguistiques sont indispensables pour construire une représentation informatique de la phase et des connaissances sur le monde de référence permettent d'effectuer les inférences indispensables à une bonne compréhension. Nous décrivons la réalisation d'un analyseur sémantique qui aborde ces deux aspects de la compréhension automatique en s' inspirant d'une théorie linguistique, la sémantique ...
Digital Convergence: The Information Revolution, 1999
U of Teesside, Feb 1, 2002
The simultaneous lightness contrast (SLC) is the condition whereby a grey patch on a dark backgro... more The simultaneous lightness contrast (SLC) is the condition whereby a grey patch on a dark background appears lighter than an equal patch on a light background. Interestingly, the lightness difference between the grey patches increases when the two backgrounds—although kept at the same luminance average—are patterned shaping what can be named the articulated-SLC. There are two main interpretations for these phenomena. The framework approach maintains that the visual system groups the luminance within a set of ...
Intellectica, 1996
Résumé/Abstract Les rapports entre le contenu sémantique et les formalismes de description sont u... more Résumé/Abstract Les rapports entre le contenu sémantique et les formalismes de description sont un problème majeur pour le développement d'une technologie informatique des objets culturels, dont le traitement automatique des textes n'est qu'une première étape. Dans cet article, en prenant comme cadre théorique la sémiotique du texte, nous passons en revue quelques problèmes classiques liés à la description du contenu sémantique. Ceci nous amène à avancer un certain nombre de directions de recherche pour aborder les ...
Seventh Intelligent Narrative Technologies Workshop, Oct 23, 2014
Virtual Environments (VE) can be used to represent complex situations. It is thus a logical evolu... more Virtual Environments (VE) can be used to represent complex situations. It is thus a logical evolution to extend them to carry out problem solving tasks as well. Intelligent Virtual Environments (IVE) integrate AI techniques with real-time 3D environments. As such, they can support interactive problem solving, provided the underlying AI techniques can produce solutions within a time frame matching that of the user's interaction. In these systems, the visual space is directly interfaced to a problem solver. Consequently, interacting with objects in the VE serves as input to the problem solver, which outputs new solutions directly as object configurations. This kind of system has many potential applications in design, configuration, situation assessment, etc. We have implemented an intelligent virtual environment based on constraint logic programming (GNU Prolog), integrated in a realtime 3D graphic environment ( Unreal Tournament TM ) using an event-based approach. More specifically, physical interaction with the virtual world objects can be translated into real-time input to the CLP solver by selecting the variables whose values have been altered by the interaction. For instance, when visualising a configuration, the user can alter the position of certain objects. This modifies the constraints involving these objects, triggering the solver on a new set of variables. Seen from the user's perspective, the 3D environment automatically reconfigures itself by reacting to his/her interactions. This form of interactive problem solving is illustrated in the figure below. We demonstrate the system behaviour on a configuration example, which uses real-world design knowledge in terms of building interior design for offices (a bank agency in our case). In this system, the domain knowledge is expressed in the form of constraints in the object's attributes, which correspond not only to geometric constraints but also semantic constraints (such as compatibility between materials, lightning, etc.) .The system will propose a first result configuration, which can be interactively refined by the user, through direct physical modification in the virtual world, triggering the computation of a new solution compatible with the user's input. Our results show that the sampling rate of object manipulation in the virtual environment is compatible with the result production of the solver (on average, the total time is less than 0.15 seconds), hence supporting interactive exploration of the solution space.
This paper presents the first prototype of a reconfigurable Virtual Environment (VE). The objecti... more This paper presents the first prototype of a reconfigurable Virtual Environment (VE). The objective of the system is to link 3D Intelligent Virtual Environments to interactive planning systems. This type of system makes possible interactive solutions where the user refines a possible configuration and enables the system to generate a complete new solution enforcing all the design constraints, previously programmed. In this first prototype we link our constraint solver with the visualization engine so that the solution produced by the constraint solver is displayed in a VE.
In this paper, we discuss technical issues in the implementation of Intelligent Virtual Environme... more In this paper, we discuss technical issues in the implementation of Intelligent Virtual Environments (IVE), taking as a starting point their requirement for a specific Knowledge Representation layer. As IVE should combine both an abstract level of representation and a concrete model of some physical world, with which the user can directly interact, there is a need for unifying principles that can describe the relations between the user, the Knowledge Representation layer and the Virtual World. We suggest that, while the user can still interact physically with the Virtual Environment, he can also exchange abstract information with the underlying Knowledge Representation layer through Natural Language, and that this information should be interpreted into corresponding state of affairs in the Virtual Environment. Conversely, there is also a need for describing the Virtual World dynamics in terms of the high-level concepts that populate the Knowledge Representation layer. After illustrating these issues in the context of "Brightown" scenarios, we discuss the use of some Artificial Intelligence techniques and formalisms in the specific context of Intelligent Virtual Environments and outline a framework for their implementation.
A fundamental issue when building a semantic lexicon for a Natural Language Understanding (NLU) s... more A fundamental issue when building a semantic lexicon for a Natural Language Understanding (NLU) system hinges on the amount of world knowledge that should be incorporated into the lexicon: the so-called dictionary-encyclopedia controversy. This issue can nd a technical solution if the NLU architecture separates a linguistic step and a domain inference step. In that case, the lexicon can contain purely linguistic elements. This requires a good descriptive linguistic theory, and an explicit link between knowledge representation and linguistic data. We hereby report preliminary results from a program embedding these principles.
In this paper, we report work in progress in the development of a novel platform for Virtual Real... more In this paper, we report work in progress in the development of a novel platform for Virtual Reality Art installations. We first introduce the notion of Alternative Reality that extends traditional VR to support the more radical experiments required by artistic productions. We subsequently introduce several technical approaches through which traditional laws of physics or more fundamental laws such as causality can be redefined within a virtual environment. We use a game engine as a development environment, which supports the redefinition of low-level physical behaviours. Changing physical or causal laws however requires the development of an explicit representation of high-level behaviour. Working with causality in particular calls for the use of representations developed for instance in qualitative physics. We illustrate the work in progress with early results from small-scale examples and finally describe the target platform onto which final installations will be developed.
Studies in Computational Intelligence, 2007
Summary. Cardiac emergency is one field in medicine widely studied and very often experienced in ... more Summary. Cardiac emergency is one field in medicine widely studied and very often experienced in reality. Furthermore, simulation and training issues on this field require sensitive and expensive techniques. In the following, we describe the development of a virtual human to be used for training applications in the field of cardiac emergencies. The system integrates Artificial Intelligence (AI) techniques for simulating medical conditions (shock states) with a realistic visual simulation of the patient in a 3D environment ...
Lecture Notes in Computer Science, 1998
Avatars and Artificial Actors in Virtual Environments can be controlled by speech, as an alternat... more Avatars and Artificial Actors in Virtual Environments can be controlled by speech, as an alternative to motion capture techniques. In this paper, we discuss some specific requirements for the successful implementation of speech-based control of guided actors. We describe our sublanguage approach to speech-based control and its associated parsing techniques, based on lexicalised grammars. After an introduction to the REALISM animation software, we report work in progress in the real-time processing of spoken commands ...
International Conference on Multimodal Interfaces and the Workshop on Machine Learning for Multimodal Interaction on - ICMI-MLMI '10, 2010
In this paper, we investigate the user's eye gaze behavior during the conversation with an intera... more In this paper, we investigate the user's eye gaze behavior during the conversation with an interactive storytelling application. We present an interactive eye gaze model for embodied conversational agents in order to improve the experience of users participating in Interactive Storytelling. The underlying narrative in which the approach was tested is based on a classical XIX th century psychological novel: Madame Bovary, by Flaubert. At various stages of the narrative, the user can address the main character or respond to her using free-style spoken natural language input, impersonating her lover. An eye tracker was connected to enable the interactive gaze model to respond to user's current gaze (i.e. looking into the virtual character's eyes or not). We conducted a study with 19 students where we compared our interactive eye gaze model with a non-interactive eye gaze model that was informed by studies of human gaze behaviors, but had no information on where the user was looking. The interactive model achieved a higher score for user ratings than the non-interactive model. In addition we analyzed the users' gaze behavior during the conversation with the virtual character.