Josiane Mothe

DocCube: multi-dimensional visualisation and exploration of large document sets

This paper presents a novel user interface that provides global visualizations of large document sets in order to help users to formulate the query that corresponds to their information needs and to access the corresponding documents. An important element of the approach we introduce is the use of concept hierarchies (CHs) in order to structure the document collection. Each CH corresponds to a facet of the documents users can be interested in. Users browse these CHs in order to specify and refine their information needs. Additionally the interface is based on OLAP principles and multi-dimensional analysis operators are provided to users in order to allow them to explore a document collection.

Nonconvex Regularizations for Feature Selection in Ranking With Sparse SVM

Feature selection in learning to rank has recently emerged as a crucial issue. Whereas several preprocessing approaches have been proposed, only a few have focused on integrating feature selection into the learning process. In this paper, we propose a general framework for feature selection in learning to rank using support vector machines with a sparse regularization term. We investigate both classical convex regularizations, such as ℓ1 or weighted ℓ1, and nonconvex regularization terms, such as log penalty, minimax concave penalty, or ℓp pseudo-norm with p<;1. Two algorithms are proposed: the first, an accelerated proximal approach for solving the convex problems, and, the second, a reweighted ℓ1 scheme to address nonconvex regularizations. We conduct intensive experiments on nine datasets from Letor 3.0 and Letor 4.0 corpora. Numerical results show that the use of nonconvex regularizations we propose leads to more sparsity in the resulting models while preserving the prediction performance. The number of features is decreased by up to a factor of 6 compared to the ℓ1 regularization. In addition, the software is publicly available on the web.

TetraFusion: information discovery on the Internet

The TetraFusion system described in this paper supports knowledge discovery from the World Wide Web by helping users perform data mining operations on sets of harvested URLs. Potential applications range from domain overviewing to science monitoring to competitive intelligence.

Combining mining and visualization tools to discover the geographic structure of a domain

Abstract Science monitoring is a core issue in the new world of business and research. Companies and institutes need to monitor the activities of their competitors, get information on the market, changing technologies or government policies. This paper presents the Tetralogie platform that is aimed at allowing a user to interactively discover trends in scientific research and communities from large textual collections that include information about geographical location. Tetralogie consists of several agents that communicate with each other on users’ demands in order to deliver results to them. Metadata and document content are extracted before being mined. Results are displayed in the form of histograms, networks and geographical maps; these complementary types of presentations increase the possibilities of analysis compared to the use of these tools separately. We illustrate the overall process through a case study of scientific literature analysis and show how the different agents can be combined to discover the structure of a domain. The system correctly predicts the country contribution to a field in future years and allows exploration of the relationships between countries.

Fusing different information retrieval systems according to query-topics: a study based on correlation in information retrieval systems and TREC topics

To evaluate Information Retrieval Systems on their effectiveness, evaluation programs such as TREC offer a rigorous methodology as well as benchmark collections. Whatever the evaluation collection used, effectiveness is generally considered globally, averaging the results over a set of information needs. As a result, the variability of system performance is hidden as the similarities and differences from one system to another are averaged. Moreover, the topics on which a given system succeeds or fails are left unknown. In this paper we propose an approach based on data analysis methods (correspondence analysis and clustering) to discover correlations between systems and to find trends in topic/system correlations. We show that it is possible to cluster topics and systems according to system performance on these topics, some system clusters being better on some topics. Finally, we propose a new method to consider complementary systems as based on their performances which can be applied for example in the case of repeated queries. We consider the system profile based on the similarity of the set of TREC topics on which systems achieve similar levels of performance. We show that this method is effective when using the TREC ad hoc collection.

Overview of ImageCLEF 2017: information extraction from images

This paper presents an overview of the ImageCLEF 2017 evaluation campaign, an event that was organized as part of the CLEF (Conference and Labs of the Evaluation Forum) labs 2017. ImageCLEF is an ongoing initiative (started in 2003) that promotes the evaluation of technologies for annotation, indexing and retrieval for providing information access to collections of images in various usage scenarios and domains. In 2017, the 15th edition of ImageCLEF, three main tasks were proposed and one pilot task: (1) a LifeLog task about searching in LifeLog data, so videos, images and other sources; (2) a caption prediction task that aims at predicting the caption of a figure from the biomedical literature based on the figure alone; (3) a tuberculosis task that aims at detecting the tuberculosis type from CT (Computed Tomography) volumes of the lung and also the drug resistance of the tuberculosis; and (4) a remote sensing pilot task that aims at predicting population density based on satellite images. The strong participation of over 150 research groups registering for the four tasks and 27 groups submitting results shows the interest in this benchmarking campaign despite the fact that all four tasks were new and had to create their own community.

Connectionist and Genetic Approaches for Information Retrieval

In the past few decades, knowledge based techniques have made an impressive contribution to intelligent information retrieval (IR). These techniques stem from research on artificial intelligence, neural networks (NN) and genetic algorithms (GA) and are used to answer three main IR tasks: information modelling, query evaluation and relevance feedback. The paper describes IR approaches based on connectionist and genetic approaches. Our goal is to take benefits of these techniques to fulfill the user information needs. More precisely a multi-layer NN, Mercure, is used to represent the document space in an associative way, to evaluate the query using spreading activation and to implement a relevance feedback process by relevance back-propagation. Another query reformulation technique is investigated which uses the GA approach. The GA generates several queries that explore different areas of the document space. Experiments and results obtained with both techniques are shown and discussed.

How many performance measures to evaluate information retrieval systems

Evaluating effectiveness of information retrieval systems is achieved by performing on a collection of documents, a search, in which a set of test queries are performed and, for each query, the list of the relevant documents. This evaluation framework also includes performance measures making it possible to control the impact of a modification of search parameters. The program trec_eval calculates a large number of measures, some being more used like the mean average precision or recall-precision curves. The motivation of our work is to compare all measures and to help the user to choose a small number of them when evaluating different information retrieval systems. In this paper, we present the study we carried out from a massive data analysis of TREC results. Relationships between the 130 measures calculated by trec_eval for individual queries are investigated, and we show that they can be clustered into homogeneous clusters.

An adaptable search engine for multimodal information retrieval

It cannot be overemphasized that changes in concepts have far more impact than new discoveries

Interactive Visual User Interfaces: A Survey

Following a short survey of input data types onwhich to construct interactive visual userinterfaces, we report on a new and recentimplementation taking concept hierarchies asinput data. The visual user interfacesexpress domain ontologies which are based onthese concept hierarchies. We detail aweb-based implementation, and show examples ofusage. An appendix surveys related systems,many of them commercial.