Laurence Capus

A case-based reasoning approach to support story summarization

The automatic generation of summaries using cases (GARUCAS) environment was designed as an intelligent system to help one learn to summarize narrative texts by means of examples within a case‐based reasoning (CBR) approach. Each example, modeled as a case, contains a conceptual representation of the initial textual state, the different steps of the summarization method, and the representation of the final textual state obtained. The CBR approach allows the environment to summarize new texts in order to produce new text summarization examples with respect to some predefined educational objectives. Within GARUCAS, this approach is used at two levels: an event level (EL) in order to identify essential elements of a story, and the clause level (CL) to make the summary more readable. The purpose of this article is to describe the GARUCAS environment and the model used to build story summarization examples and summarize new texts. This model is based on important psycholinguistic work concerning event and narrative structures and text revision rules. An experiment was conducted with 12 short stories. The GARUCAS environment can classify the stories according to their structure analogy and reuse the summarization method of the most similar text. Such an approach can be reused for any kind of texts or summary types.

A Learner Model for Learning-by-Example Context

Nowadays learning environments put more and more accent on the intelligence of the system. The intelligence of a learning environment is largely attributed to its ability of adapting to a specific learner during the learning process. The adaptation depends on individual learners knowledge of the subject to be learned, and other relevant characteristics of the learner. The knowledge and the relevant information about the learner are maintained in the learner model. A learner model can be defined as structured information about the learning process; and this structure contains some values of the learners characteristics. This paper proposes a new learner model, which is based on the consideration of what is appropriate to the learning-by-example context. The model records five categories of information about the learner: personal data, learners characteristics, learning state, learners interactions with the system, and learners knowledge. This model is being integrated in Sphinx, an educational environment based on learning by means of examples.

Insertion of Ontological Knowledge to Improve Automatic Summarization Extraction Methods

The vast availability of information sources has created a need for research on automatic summarization. Current methods perform either by extraction or abstraction. The extraction methods are interesting, because they are robust and independent of the language used. An extractive summary is obtained by selecting sentences of the original source based on information content. This selection can be automated using a classification function induced by a machine learning algorithm. This function classifies sentences into two groups: important or non-important. The important sentences then form the summary. But, the efficiency of this function directly depends on the used training set to induce it. This paper proposes an original way of optimizing this training set by inserting lexemes obtained from ontological knowledge bases. The training set optimized is reinforced by ontological knowledge. An experiment with four machine learning algorithms was made to validate this proposition. The improvement achieved is clearly significant for each of these algorithms.

Evaluation of Efficiency of Linear Techniques to Optimize Attribute Space in Machine Learning: Relevant Results for Extractive Methods of Summarizing

One major challenge in the field of machine learning, especially in classification problems, is to optimize the attribute space in order to obtain a classification function, which will be used to discriminate future items. Several approaches to optimize the attribute space can be used: some of them select the most relevant attributes and the other ones extract certain attributes to create a new smaller set of variables. These classification approaches have recently been implemented in the automatic summarization process with promising results. This paper enriches these first results with another new experiment. Five well-known linear methods were exploited to optimize the attribute space in an original manner on a corpus of 1250 text documents. These methods, used in data clustering and unsupervised machine learning, allow either attribute selection (Singular Value Decomposition, K-Means, Kohonen Neural Networks) or new attribute extraction (Principal Component Analysis, Factor Analysis). After having applied these methods to optimize attribute space, the validation phase was focused on the discrimination power of the obtained classification function. For that, six techniques of machine learning were used to abduce the classification function. Its performance was evaluated with the metric Fmesure and ROC curves. The results show that the application of the five chosen linear methods for optimizing attribute space in the automatic summarization process by extraction is relevant. They also show which machine learning technique is preferable to use with each linear method to obtain a better efficiency.

Une interface plus intelligente pour SPHINX, un système d'aprentissage humain à partir d'exemples

Sphinx is a computer environment to help human learning by means of examples in an introductory course of artificial intelligence. This interactive Web environment offers learners specific tools to help them explain the solved problems, alone or in collaboration with peers. Using Sphinx, learners can obtain from their peers or their teachers feedback on the produced explanations. In order to improve the interface between the learners and Sphinx, we added some functionalities for a better use of Sphinx by learners: personalization of the interface by the learner, selection of the examples by the system according to the learners knowledge, intelligent guidance and advices by Sphinx. These new functions have been implemented and are presently under evaluation.

Improving Traffic Lights System Management by Translating Decisions of Traffic Officer

Coordination of traffic signal timing systems has significant impacts on traffic congestion, waiting time, risks of accidents, and unnecessary fuel consumption. Actually, systems of traffic light’s programming involve complex calculations especially to tackle problematic situations in real time. Another way of doing is to manage traffic flow by traffic officers. Despite the limitation of short-term retention of human brain to few elements, human being can make decisions in case of system malfunction or during special events. The human strategy as that of the traffic officers is simple and is based on common sense. This paper explains how to implement this strategy and gives some results obtained. The simulation is performed with the open-source traffic simulation software, simulation of urban mobility (SUMO). The preliminary simulation results are promising for the continuation of this research. The observation of patterns could bring to propose an intelligent system more efficient that reuses similar cases to save time.

Including Personality Traits, Inferred from Social Networks, in Building Next Generation of AEHS

User profile inference on online social networks is a promising way for building recommender and adaptive systems. In the context of adaptive learning systems, user models are still constructed by means of classical techniques such as questionnaires. Those are too time-consuming and present a risk of dissuading learners to use the system. This paper explores the feasibility of learner modeling based on a proposed set of features extracted and inferred from social networks, according to the IMS-LIP specification. A suitable general architecture of an AEHS is presented, whose adaptation combines three distinct aspects: Felder and Silverman learning style, knowledge level and personality traits. This latter is a novel adaptation criterion, it is an interesting user feature to be incorporated in user models, a feature that is not yet considered by existing AEHS. However, adapting such systems to personality traits contributes to achieving a better adaptation by varying learning approaches, integrating collaboration and adapting feedback. The aim of this paper is to show how this contribution is doable through the proposed framework.

VENCE: A new machine learning method enhanced by ontological knowledge to extract summaries

Obtaining extractive summaries by using functions induced from a training set continue to be a great challenge in the domain of the automatic text summary. This paper presents the VENCE method based on this approach and improves the quality of the abduced functions, using semantic relations of the words (attributes) of the training set that are fetched from a ontology to be inserted in this set. The choice of this training set is reinforced with the optimization of the space of attributes by means of statistical techniques, as well as with the introduction of the Jaccard index, calculated from considering a manual summary that is extracted from the corpus of the chosen documents. The VENCE method is explained in details as well as the different experiments conducted to propose an optimal process. Its application to a text document corpus highlighted its efficiency. The results obtained are very satisfactory for the assessment of discriminating power of the abduced classification function as well as for the quality of summaries produced.

A Cognitive Model for Automatic Narrative Summarization in a Self-Educational System

The use of examples may be a good strategy to learn or to improve one’s abilities in a particular field; but the number of educational systems using examples explicitly remains small. The goal of our GARUCAS (GenerAtor of summaRies Using CASes) project is to build an educational system to help users learn text summarization by means of examples. It uses case-based reasoning to build new summaries from old ones. By observing the expert module producing a summary, the system user will learn how to summarize. Thus, the system needs examples of text summarization in order to show how summaries are produced and also to reuse in summarizing other texts. We began with 12 examples of summarization of simple narrative texts.