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

No abstract

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 F mesure 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.