Computing with perceptions
The figure shows the main components of a computational system to generate relevant linguistic descriptions of phenomena in different situation types. The development of this type of systems is a multi-disciplinary project where:
The development of this type of systems is a multi-disciplinary project where:
- The Data acquisition module depends on each application and it could be rather complex. For instance, in a system capable of describing the traffic in a road, the data acquisition module will include the image analysis and pattern recognition tasks necessary to identify the presence of vehicles in the road and compute their velocity.
- In our approach, the Linguistic Model is formed by an ontology with elements belonging to three different domains: (a) concepts belonging to the application itself, (b) concepts developed in the context of Computational Theory of Perceptions and (c) elements belonging to the language itself, e.g.: English or Spanish, taken from the Systemic Functional Linguistics.
- The Interpreter module uses the Linguistic Model to recognize the meaning of input data and then to generate relevant linguistic utterances.
According with this general architecture, our project deals with the following goals:
- Extend the Computational Theory of Perceptions by developing computational models of phenomena with capacity for representing imprecision with different levels of granularity.
- Apply Systemic Functional Linguistic to represent the particular way of using natural language in each specific application field.
- Design a computational model of emotions that allow us generating texts with emotional content.
- Develop ontologies to represent the domain of knowledge and language in each application domain.
- When needed, design and develop the data acquisition techniques required in each type of application.
The project plan is organized in several interrelated research lines that are listed as follows:
- Definition of the basic concepts, including the architecture and methodology required to build this type of systems. This research line contributes to the development of Computational Theory of Perceptions with new concepts and procedures. An specific goal is to combine elements of this theory with Systemic Functional Linguistic and resources obtained from the, so called, Semantic Web.
- Application 1: Linguistic reporting in training simulators. We collaborate with the company ESM (www.esm.es) to generate linguistic reports about the impact of new devices in driving safety (e.g., smartphone and GPS).
- Application 2: Generation of linguistic advices to improve the efficiency in the consumption of energy. We collaborate with HCenergia (www.hcenergia.es) to generate linguistic advices that will be included in the client’s invoices.
- Application 3: Linguistic description of images. We are currently working in the description of traffic behavior. Also, we collaborate with INTA (www.inta.es) in the linguistic descriptions of images of the surface of Mars planet.
- Computational model of emotions. We aim to develop a computational model of the temporal evolution of emotions of an avatar. This will allow us to generate linguistic expresions with emotional content.
- Linguistic modeling of human activities. We have used accelerometers and WI-FI signal to recognize human and describe activities. Currently, we use smatphones to monitor human gestures, postures and localization. We model the human gait to recognize a person and also to analyze the gait quality.
- Design and development of data acquisition devices. We have developed several devices to measure accelerations, light, temperature and skin conductivity. As a result of this activity we have registered several patents.
The publications and details about the projects are available in the corresponding section of this web site.