Hakim Hacid

A New Context-Aware Measure for Semantic Distance Using a Taxonomy and a Text Corpus

Having a reliable semantic similarity measure between words/concepts can have major effect in many fields like information retrieval and information integration. A major lack in the existing semantic similarity measures is that no one takes into account the actual context or the considered domain. However, two concepts similar in one context may appear completely unrelated in another context. In this paper, we present a new context-based semantic distance. Then, we propose to combine it with classical approaches dealing with taxonomies and corpora. Our correlation ratio of 0.89 with human judgments on a set of words pairs led our approach to outperform all the other approaches.

A multisource context-dependent semantic distance between concepts

A major lack in the existing semantic similarity methods is that no one takes into account the context or the considered domain. However, two concepts similar in one context may appear completely unrelated in another context. In this paper, our first-level approach is context-dependent. We present a new method that computes semantic similarity in taxonomies by considering the context pattern of the text corpus. In addition, since taxonomies and corpora are interesting resources and each one has its strengths and weaknesses, we propose to combine similarity methods in our second-level multi-source approach. The performed experiments showed that our approach outperforms all the existing approaches.

Neighborhood graphs for semi-automatic annotation of large image databases

Images annotation is the main tool for associating a semantic to an image. In this article we are interested in the semi-automatic annotation of images data. Indeed, with the great mass of data managed throughout the world and especially with the Web, the manual annotation of these images is almost impossible. We propose an approach based on neighborhood graphs offering several possibilities: content-based retrieval, key-words based interrogation, and the annotation which concerns us in this article. The approach we are proposing offers, as the experiments section shows it, very interesting annotation results while satisfying the scalability criteria which is a very significant point in this context where the mass of data is very important.

Mining semantic distance between corpus terms

In this paper, we face two problems in classical semantic similarity measures. Firstly, the context-dependency problem in knowledge-base measures since no one takes into account the context of the target domain. That is, a multisource context-dependent approach is presented. Secondly, the coverage problem with these measures since similarities can only be calculated between concepts included in a taxonomy. Moreover, pure corpus-based measures are still way from achieving performance reached by knowledge based measures. We present a more complex corpus-based approach using a taxonomy and data mining techniques in order to compute semantic distances between terms uncovered by the taxonomy. Experiments made show clearly the effectiveness of both proposed approaches.

Exploring Validity Indices for Clustering Textual Data

The goal of any clustering algorithm producing flat partitions of data, is to find both the optimal clustering solution and the optimal number of clusters. One natural way to reach this goal without the need for parameters, is to involve a validity index in a clustering process, which can lead to an objective selection of the optimal number of clusters. In this chapter, we provide two main contributions. Firstly, since validity indices have been mostly studied in a two or three-dimensionnal datasets, we have chosen to evaluate them in a real-world applications, document and word clustering. Secondly, we propose a new context-aware method that aims at enhancing the validity indices usage as stopping criteria in agglomerative algorithms. Experimental results show that the method is a step-forward in using, with more reliability, validity indices as stopping criteria.

Using semantic distance in a content-based heterogeneous information retrieval system

This paper brings two contributions in relation with the semantic heterogeneous (documents composed of texts and images) information retrieval: (1) A new context-based semantic distance measure for textual data, and (2) an IR system providing a conceptual and an automatic indexing of documents by considering their heterogeneous content using a domain specific ontology. The proposed semantic distance measure is used in order to automatically fuzzify our domain ontology. The two proposals are evaluated and very interesting results were obtained. Using our semantic distance measure, we obtained a correlation ratio of 0.89 with human judgments on a set of words pairs which led our measure to outperform all the other measures. Preliminary combination results obtained on a specialized corpus of web pages are also reported.

On determining the optimal partition in agglomerative clustering of documents

The goal of any clustering algorithm producing flat partitions of data, is to find both the optimal clustering solution and the optimal number of clusters. One natural way to reach this goal without the need for parameters, is to involve a validity index in a clustering process, which can lead to an objective selection of the optimal number of clusters. In this paper, we provide an evaluation of the major relative indices involving them in an agglomerative clustering algorithm for documents. The evaluation seeks the indices ability to identify both the optimal solution and the optimal number of clusters. Then, we propose a new context-aware method that aims at enhancing the validity indices usage as stopping criteria in agglomerative algorithms. Experimental results show that the method is a step-forward in using, with more reliability, validity indices as stopping criteria.

CSDTM a cost sensitive decision tree based method

Making a decision has often many results and repercussions. These results dont have the same importance according to the considered phenomenon. This situation can be described by the introduction of the cost concept in the learning process. In this article, we propose a method able to integrate the costs in the automatic learning process. We focus our work on the misclassification cost and we use decision trees as a supervised learning technique. Promising results are obtained using the proposed method.