Abstract Artificial Intelligence encompasses a range of technologies that replicate human-like cognitive abilities through computer systems, enabling the execution of tasks associated with intelligent beings. A prominent way to achieve this is machine learning (ML), which optimizes system performance by employing learning algorithms to create models based on data and its inherent patterns. Today, a multitude of ML models exist having diverse characteristics, including the algorithm type, training dataset, and resultant performance. Such diversity complicates the selection of an appropriate model for a specific use case, answering user demands. This paper presents an approach for ML models retrieval based on the matching between user inputs and ML models criteria, all described in a semantic ML ontology named SML model (Semantic Machine Learning model), which facilitates the process of ML models selection. Our approach is based on similarities measures that we tested and experimented to score the ML models and retrieve the ones matching, at best, user inputs.

Recent advances in sensor technology and information processing have allowed connected environments to impact various application domains. In order to detect events in these environments, existing works rely on the sensed data. However, these works are not re-usable since they statically define the targeted events (i.e., the definitions are hard to modify when needed). Here, we present a generic framework for event detection composed of (i) a representation of the environment; (ii) an event detection mechanism; and (iii) an Event Query Language (EQL) for user/framework interaction. This paper focuses on detailing the EQL which allows the definition of the data model components, handles instances of each component, protects the security/privacy of data/users, and defines/detects events. We also propose a query optimizer in order to handle the dynamicity of the environment and spatial/temporal constraints. We finally illustrate the EQL and conclude the paper with some future works.

Semantic web techniques (e.g., ontologies) have been recently adopted for sensor network modeling. However, existing works do not fully address these challenges: (i) representing different sensor types (e.g., mobile/static sensors) to enrich the network with different data and ensure better coverage; (ii) representing a variety of platforms (e.g., environments, devices) for sensor deployment, thus, integrating new components (e.g., mobile phones); (iii) representing the diverse data (scalar/multimedia) needed for various applications (e.g., event detection); and (iv) proposing a generic model to allow re-usability in various application domains. In this paper, we propose HSSN, an ontology that extends the Semantic Sensor Network (SSN) ontology which is already re-usable and considers various platforms. We extend the representation of sensors, sensed data, and deployment environments to cope with these challenges. We evaluate the consistency, accuracy, clarity, and performance of HSSN.

In the context of RDF document matching/integration, the datatype information, which is related to literal objects, is an important aspect to be analyzed in order to better determine similar RDF documents. In this paper, we present an R DF D atatype in F erring F ramework, called RDF-F, which provides two independent datatype inference processes: 1) a four-step process consisting of (i) a predicate information analysis (i.e., deduce the datatype from existing range property), (ii) an analysis of the object value itself by a pattern-matching process (i.e., recognize the object lexical space), (iii) a semantic analysis of the predicate name and its context, and (iv) generalization of Numeric and Binary datatypes to ensure the integration; and 2) a non-ambiguous lexical-space-matching process, where literal values are inferred by the modification of their representation, following new lexical spaces. We evaluated the performance and the accuracy of both processes with datasets from DBpedia. Results show that the execution time of both indicators is linear and their accuracy can increase up to 97.10 and 99.30%, respectively.

Multi-Agent Systems (MASs) have received much attention in recent years because of their advantages on modeling complex distributed systems, such Digital Ecosystems (DESs). Many existing modeling languages that support the design of such systems are based on ontologies to assist the representation of agents knowledge. However, in the context of DESs, there is still a need for more general conceptual models to represent the specific characteristics of DESs in terms of win-win interaction, engagement, equilibrium, and self-organization. Then, concepts such behavior, roles, rules, and environment are needed. This paper describes an ontology-based approach by proposing MAS2DES-Onto, as the conceptual model, which considers the essential static and dynamic aspects of MASs by a clear representation of their concepts and relationships to support the design and development of DESs. To validate and conduct experimental tests, we integrate MAS2DES-Onto into a framework to automatically generate MAS-based DESs. Results show the efficiency and effectiveness of our approach.

International audience

Existing XML query algebras are not fully appropriate to retrieve RSS news items mainly due to three reasons: 1) RSS is text rich and its content is dependent on the wording and verbification of the author, thus semantic aware operators are needed; 2) news items are dynamic and consequently time oriented retrieval is needed; 3) a news item may evolve through time, or overlap with other news items and hence identifying relationships between items is also needed. In this paper, we aim to solve these issues by providing a dedicated RSS algebra based on semantic-aware operators that consider RSS characteristics. The provided operators are application domain specific and can be tuned according to the user preferences. We also provide a set of query rewriting and equivalence rules that would be used during query simplification and optimization. In addition and in order to validate our proposal, we present here our prototype that allows a user to formulate RSS query using our operators.

XML data flow has reached beyond the world of computer science and has spread to other areas such as data communication, e-commerce and instant messaging. Therefore, manipulating this data by non expert programmers is becoming imperative. On one hand, Mashups have emerged a few years ago, providing users with visual tools for web data manipulation but not necessarily XML specific. Mashups have been leaning towards functional composition but no formal languages have yet been defined. On the other hand, visual languages for XML have been emerging since the standardization of XML, and mostly relying on querying XML data for extraction or structure transformations. These languages are mainly based on existing textual XML languages, have limited expressiveness and do not provide non expert programmers with means to manipulate XML data. In this paper, we define a generic visual language called XCDL based on Colored Petri Nets allowing non expert programmers to compose manipulation operations. The language is adapted to XML, providing users with means to compose XML oriented operations. The language core syntax is presented here along with an implemented prototype based on it.

In this paper, we describe our Multimedia Context based Security Engine (MCSE) which is a Java Based Prototype able to integrate multimedia context in order to enforce access control policies. The prototype provides supervised access to a database containing sensitive viral images.

Since the last two decades, image database management has been practiced using different image representation methods. In the literature, images are represented using two paradigms: the metadata-based and the content-based representations. Image retrieval using the metadata is done using the traditional database operations. However, image retrieval by its low-level features requires similarity-based operations. Practice has shown that both types of operations are needed for an efficient image database management system. Particularly in medical image databases, such a mixed form of retrieval is very important. We first present a global image data model that supports both metadata and low-level descriptions of images. We illustrate our work with real examples in the medical domain. Then, using an original image data repository model, we show how relational and similarity-based operations can be integrated. Both image and salient object are considered in our model. A prototype called MIMS (medical image management system) has been realized to validate the main aspects of our approach.