Jean-Pierre Pécuchet

Context and Keyword Extraction in Plain Text Using a Graph Representation

Document indexation is an essential task achieved by archivists or automatic indexing tools. To retrieve relevant documents to a query, keywords describing this document have to be carefully chosen. Archivists have to find out the right topic of a document before starting to extract the keywords. For an archivist indexing specialized documents, experience plays an important role. But indexing documents on different topics is much harder. This article proposes an innovative method for an indexing support system. This system takes as input an ontology and a plain text document and provides as output contextualized keywords of the document. The method has been evaluated by exploiting Wikipedias category links as a termino-ontological resources.

Conceptual Indexing of Documents Using Wikipedia

This paper presents an indexing support system that suggests for librarians a set of topics and keywords relevant to a pedagogical document. Our method of document indexing uses the Wikipedia category network as a conceptual taxonomy. A directed acyclic graph is built for each document by mapping terms (one or more words) to a concept in the Wikipedia category network. Properties of the graph are used to weight these concepts. This allows the system to extract so called important concepts from the graph and to disambiguate terms of the document. According to these concepts, topics and keywords are proposed. This method has been evaluated by the librarians on a corpus of french pedagogical documents.

LEARNING SVM WITH COMPLEX MULTIPLE KERNELS EVOLVED BY GENETIC PROGRAMMING

Classic kernel-based classifiers use only a single kernel, but the real-world applications have emphasized the need to consider a combination of kernels — also known as a multiple kernel (MK) — in order to boost the classification accuracy by adapting better to the characteristics of the data. Our purpose is to automatically design a complex multiple kernel by evolutionary means. In order to achieve this purpose we propose a hybrid model that combines a Genetic Programming (GP) algorithm and a kernel-based Support Vector Machine (SVM) classifier. In our model, each GP chromosome is a tree that encodes the mathematical expression of a multiple kernel. The evolutionary search process of the optimal MK is guided by the fitness function (or efficiency) of each possible MK. The complex multiple kernels which are evolved in this manner (eCMKs) are compared to several classic simple kernels (SKs), to a convex linear multiple kernel (cLMK) and to an evolutionary linear multiple kernel (eLMK) on several real-world data sets from UCI repository. The numerical experiments show that the SVM involving the evolutionary complex multiple kernels perform better than the classic simple kernels. Moreover, on the considered data sets, the new multiple kernels outperform both the cLMK and eLMK — linear multiple kernels. These results emphasize the fact that the SVM algorithm requires a combination of kernels more complex than a linear one in order to boost its performance.

Distributed Multiagent Simulation on P2P Architecture

In this paper we introduce distributed multiagent simulation and recall some solutions based on HLA protocol. We make a short presentation of peer-to- peer systems. We highlight the architecture of a distributed multiagent system based on a peer-to-peer model. We present some preliminary experimental results to show the performance of this architecture running a distributed version of ants model. We compare this implementation with a non-distributed ants version and with a distributed version on Client/Server architecture.In this paper we introduce distributed multiagent simulation and recall some solutions based on HLA protocol. We make a short presentation of peer-topeer systems. We highlight the architecture of a distributed multiagent system based on a peer-to-peer model. We present some preliminary experimental results to show the performance of this architecture running a distributed version of ants model. We compare this implementation with a non-distributed ants version and with a distributed version on Client/Server architecture.

Empirical Specification of Dialogue Games for an Interactive Agent

This article addresses the challenge of designing the communicative behaviour of an agent interacting with humans. We present a data-driven methodology based on the production of a matrix representation of a corpus from which we extract dialogue patterns. These patterns reflect the minimal units of interaction which turn out to be very attractive for dialogue modelling. We present a framework to specify dialogue games from these patterns based on the notion of social commitments. We exemplify the specification of dialogue games by implementing all the steps of our methodology on a task-oriented corpus. The produced games are validated by showing that they appropriately describe the patterns appearing in a reference corpus.

Fusion of Smile, Valence and NGram Features for Automatic Affect Detection

This paper addresses the problem of feature fusion between smile, as a visual feature, and text, as a transcription result. The influence of smile over semantic data has been considered before, without investigating multiple approaches for the fusion. This problem is multi-modal, which makes it more difficult. The goal of this article is to investigate how this fusion could increase the current interactivity of a dialogue system by boosting the automatic detection rate of the sentiments expressed by a human user. There are two original propositions in our approach. The first lies in the use of a segmented detection for text data, rather than predicting a single label for every document (video). Second, this paper studies the importance of several features in the process of multi-modal fusion. Our approach uses basic features, such as NGrams, Smile Presence or Valence to find the best fusion approach. Moreover, we test a two level classification approach, using a SVM.

Document Indexing and Retrieval Using Wikipedia

This paper introduces a framework to perform conceptual indexing and retrieval of documents. It uses a graph composed of terms from a text and elements of the Wikipedia Category Network. Conceptual indexing consists in finding the relevant Wikipedia articles and categories that can be used to describe the text. Conceptual retrieval consists in using these articles and categories to return the relevant documents for a user query. A proof-of-concept prototype is finally presented.

PHAC: an environment for distributed collaborative applications on P2P networks

Using Peer-to-Peer networks is a way to distribute large scale scientific problems. But the P2P networks are very heterogeneous, highly dynamic and volatile. The objective of our work is to offer one P2P network based on high availability. We propose our PHAC framework to improve the dependability of applications. And we present the first results with case π computing.

Design and generation of collective educational activities

This paper gives a general presentation of a method for designing, scripting and automatically generating collectives educational activities (CEAs) for distance learning. The CEAs are generated in XML (Extensible Markup Language) and can be integrated in a plate form, as a service. We first present a model which describes the content aggregation. Then, we present briefly the methodology for designing, scripting and generating the CEAs. Finally, we present the tools helping the teachers to design, build and generate the CEAs.

Using Concept Maps for Enhancing Adaptation Processes in Declarative Knowledge Learning

This article deals with adaptation techniques in the field of declarative knowledge learning. After explaining how concepts can be represented, it introduces a learner evaluation technique based on a concept maps analysis. The way an epistemic learner model can be made from this evaluation is then proposed. Finally, adaptation techniques based on this model are presented. With this method, different adaptation schemes can be applied to the document depending on the learner’s errors.