Catálogo de publicaciones - libros

Incomplete relevance judgment has become a norm for the evaluation of some major information retrieval evaluation events such as TREC, but its effect on some system measures has not been well understood. In this paper, we evaluate four system measures, namely mean average precision, R-precision, normalized average precision over all documents, and normalized discount cumulative gain, under incomplete relevance judgment. Among them, the measure of normalized average precision over all documents is introduced, and both mean average precision and R-precision are generalized for graded relevance judgment. These four measures have a common characteristic: complete relevance judgment is required for the calculation of their accurate values. We empirically investigate these measures through extensive experimentation of TREC data and aim to find the effect of incomplete relevance judgment on them. From these experiments, we conclude that incomplete relevance judgment affects all these four measures’ values significantly. When using the pooling method in TREC, the more incomplete the relevance judgment is, the higher the values of all these measures usually become. We also conclude that mean average precision is the most sensitive but least reliable measure, normalized discount cumulative gain and normalized average precision over all documents are the most reliable but least sensitive measures, while R-precision is in the middle.

The steady increase of information on WWW, digital library, portal, database and local intranet, gave rise to the development of several methods to help user in Information Retrieval, information organization and browsing. Clustering algorithms are of crucial importance when there are no labels associated to textual information or documents. The aim of clustering algorithms, in the text mining domain, is to group documents concerning with the same topic into the same cluster, producing a flat or hierarchical structure of clusters. In this paper we present a Knowledge Discovery System for document processing and clustering. The clustering algorithm implemented in this system, called Induced Bisecting k-Means, outperforms the Standard Bisecting k-Means and is particularly suitable for on line applications when computational efficiency is a crucial aspect.

This paper discusses two kinds of search strategies supported by the , an annotation service that can be used by different . The first strategy concerns the search and retrieval of annotations, considered as stand-alone documents; while, the second one regards how to exploit annotations in order to search and retrieve annotated documents which are relevant for a user query. This paper describes the proposed search strategies in the light of the architectural design choices needed to support them.

Information seekers using the usual search techniques and engines are delighted by the sheer power of the technology at their command – speed, quantity. Upon closer inspection of the results, and reflection upon the next stages of the information seeking knowledge work, users are typically overwhelmed, and frustrated. We propose a partial solution by focusing on the query formulation aspect of the information seeking problem. First we introduce our version of a semantic analysis algorithm, named Normalised Word Vector, and explain its application in assisted query formulation. Secondly we introduce our ideas of supporting query refinement via Dynamic Ontological Filtering.

Nuances in natural languages can be useful to effectively describe preferences and constraints over a complex and few formalized domain. In this paper we describe the architecture of a query answering system for the domain of motor racing which uses fuzzy logic and domain knowledge in order to carry out searches dealing with vague expression, either as search constraints or as relationship between entities attribute values.

The paper presents a fuzzy database management system, and a fuzzy method for dominant colour description of images, on which an image retrieval system is built. The paper shows the suitability of the fuzzy database management system for this kind of applications when the images are characterized by fuzzy data. The synergy of these two introduced components, improves traditional image retrieval systems in three aspects: natural and automatic image description, a natural and easy query language, and high performance in query resolution.

A typical content-based audio management system deals with three aspects namely audio segmentation and classification, audio analysis, and content-based retrieval of audio. In this paper, we integrate the three aspects of content-based audio management into a single framework and propose an efficient method for flexible querying and browsing of auditory data. More specifically, we utilize two robust feature sets namely MPEG-7 Audio Spectrum Flatness (ASF) and Mel Frequency Cepstral Coefficients (MFCC) as the underlying features in order to improve the content-based retrieval accuracy, since both features have some advantages for distinct types of audio (e.g., music and speech). The proposed system provides a wide range of opportunities to query and browse an audio data by content, such as querying and browsing for a chorus section, sound effects, and query-by-example. In addition, the clients can express their queries in the form of , , and k-, which are particularly significant in the multimedia domain.

In this paper, the theoretical aspects of calculating the similarity between sets, and its generalizations multisets, fuzzy sets and fuzzy multisets, is presented. Afterwards, this theory is applied to enhance the facilities for accessing a multimedia system, namely when searching for correspondence between multi-valued attributes, which are coupled with a thesaurus. Furthermore, to allow flexibility in this search, thesauri with similarities defined between the thesaurus terms are considered. As a possible application, the DEKKMMA project is introduced, a project about an audio archive of African music.

Research in Cooperative Query answering is triggered by the observation that users are often not able to correctly formulate queries to databases that return the intended result. Due to a lack of knowledge of the contents and the structure of a database, users will often only be able to provide very broad queries. Existing methods for automatically refining such queries based on user profiles often overshoot the target resulting in queries that do not return any answer. In this paper, we investigate methods for automatically relaxing such over-constraint queries based on domain knowledge and user preferences. We describe a framework for information access that combines query refinement and relaxation in order to provide robust, personalized access to heterogeneous RDF data as well as an implementation in terms of rewriting rules and explain its application in the context of e-learning systems.

We present Meeting Miner, a multimodal meeting browser for navigating recordings of online text and speech collaborative meetings. Meetings are recorded through a collaborative writing environment specially designed to capture participants activities. This information, usually lost in common recordings of multimodal meetings, offers novel possibilities for indexing, navigation and information retrieval in archived meetings. Meeting Miner uses temporal information from the logs of actions captured on self-contained information items (paragraphs of text) to uncover potential information links between these semantic data units. A novel space-based action navigation scheme is presented. Keywords and topic search as well as more advanced queries can be performed by the system. We illustrate the system navigation modalities with several browsing examples.