Yves Chiaramella

A prototype of an intelligent system for information retrieval: IOTA

Abstract Recent results in artificial intelligence research are of prime interest in various fields of computer science; in particular we think information retrieval may benefit from significant advances in this approach. Expert systems seem to be valuable tools for components of information retrieval systems related to semantic inference. The query component is the one we consider in this paper. IOTA is the name of the resulting prototype presented here, which is our first step toward what we call an intelligent system for information retrieval . After explaining what we mean by this concept and presenting current studies in the field, the presentation of IOTA begins with the architecture problem, that is, how to put together a declarative component, such as an expert system, and a procedural component, such as an information retrieval system. Then we detail our proposed solution, which is based on a procedural expert system acting as the general scheduler of the entire query processing. The main steps of natural language query processing are then described according to the order in which they are processed, from the initial parsing of the query to the evaluation of the answer. The distinction between expert tasks and nonexpert tasks is emphasized. The paper ends with experimental results obtained from a technical corpus, and a conclusion about current and future developments.

Information retrieval and structured documents

Standard Information Retrieval considers documents as atomic units of information that are indexed and retrieved as a whole. Modern evolution of document design and storage have since a long time introduced more elaborate representations of documents; standards such as SGML, then HTML and now XML are of course major contributions in this domain. These standards underly today evolutions towards modern electronic documents. In this context, retrieving structured documents refers to index and retrieve information according to a given structure of documents. This means that documents are no longer considered as atomic entities, but as aggregates of interrelated objects that can be retrieved separately: given a retrieval query, one may retrieve the set of document components that are most relevant to this query.In this chapter we shall first emphasise some aspects which, in our opinion, relate explicit use of document structure to interactive retrieval performances, such as efficiency while browsing or querying information. In a second step we shall investigate two classes of implementation approaches dealing with indexing and retrieving structured documents: passage retrieval and explicit use of hierarchical structures of documents.

A retrieval model based on an extended modal logic and its application to the RIME experimental approach

This paper focuses on the query processing module of RIME, an experimental prototype of an intelligent information retrieval system designed to manage high-precision queries on a corpus of medical reports. Though highly specific this particular corpus is representative of an important class of applications: information retrieval among full-text specialized documents which constitute critical sources of information in several organizations (medicine, law, space industry…). This experience allowed us to design and implement an elaborate model for the semantic content of the documents which is an extension of the Conceptual Dependency approach. The underlying retrieval model is inspired from the Logic model proposed by C.J. Van Rijsbergen, which has been considerably refined using an Extended Modal Logic. After presenting the context of the RIME project, we briefly describe the models designed for the internal representation of medical reports and queries. The main part of the paper is then devoted to the retrieval model and its application to the query processing module of RIME which has a natural language interface. Processing a query involves two main phases: the interpretation which transforms the natural language query into a search expression, and the evaluation phases which retrieves the corresponding medical reports. We focus here on the evaluation phases and show its relationship with the underlying retrieval model. Evaluations from practical experiments are also given, along with indications about current developments of the project.

IOTA: a full text information retrieval system

IOTA is a prototype of an Information Retrieval System which can manage a corpus made of highly structured, full text documents. The first version presented here has intelligent capabilities related to heuristic pattern matching procedures for processing natural language queries, which involve an automatically built thesaurus. The paper emphazises the overall principles of query processing and gives hints about the underlying techniques used while constructing the thesaurus and automatically indexing highly structured documents.

Indexing medical reports in a multimedia environment: the RIME experimental approach

This paper focuses on the RIME system aimed to the indexing of medical reports in a multimedia environment. This particular application is viewed as representative of a large set of still unanswered needs of large communities of users: domain experts dealing with on-line specialized documentation such as software engineers, medical specialists and so on. In this application textual information appears as an interesting media to access related pictures in the data base. After the presentation of the application and a study of the particular corpus involved we define a semantic model for the documents which is based on a Conceptual Language. Then we detail the indexing process and its various linguistic components which perform the translation of every medical report according to this semantic model.

A model for weighting image objects in home photographs

The paper presents a contribution to image indexing consisting in a weighting model for visible objects -- or image objects -- in home photographs. To improve its effectiveness this weighting model has been designed according to human perception criteria about what is estimated as important in photographs. Four basic hypotheses related to human perception are presented, and their validity is estimated as compared to actual observations from a user study. Finally a formal definition of this weighting model is presented and its consistence with the user study is evaluated.

A relational vector space model using an advanced weighting scheme for image retrieval

In this paper, we lay out a relational approach for indexing and retrieving photographs from a collection. The increase of digital image acquisition devices, combined with the growth of the World Wide Web, requires the development of information retrieval (IR) models and systems that provide fast access to images searched by users in databases. The aim of our work is to develop an IR model suited to images, integrating rich semantics for representing this visual data and user queries, which can also be applied to large corpora. Our proposal merges the vector space model of IR - widely tested in textual IR - with the conceptual graph (CG) formalism, based on the use of star graphs (i.e. elementary CGs made up of a single relation connected to some concepts representing image objects). A novel weighting scheme for star graphs, based on image objects size, position, and image heterogeneity is outlined. We show that integrating relations into the vector space model through star graphs increases the systems precision, and that the results are comparable to those from graph projection systems, and also that they shorten processing time for user queries.

Comparing image segmentation algorithms for Content Based Image Retrieval Systems

This article discusses how to compare different image segmentation algorithms parameters in order to choose the most optimal algorithm parameters for a specific task(s) (e.g. feature extraction, spatial reasoning, topological analysis). Our method of comparison lets the user decide which segmentation algorithm/parameters are the most suitable for his system, this decision is obtained according to three indicators: the amount of relevant visual information as well as the noise in image regions, the average number of regions per images and the average number of regions per object.

Towards a fast precision-oriented image retrieval system

Expressive indexing languages are called for to obtain faithful representations of multimedia documents. They become a decisive factor to alleviate the noise problem, which is now often encountered with commercial search engines. In this case, noise consists in a great number of non-relevant answers to user’s queries, which compromises interrogation efficiency. The use of complex formalisms to handle the problem is not straight-forward, as one of their general characteristics is to have high, sometimes unacceptable computational costs. The RELIEF image retrieval system [OP98] is the result of research conducted for several years at the IMAG research institute of the University of Grenoble. It is proposed as a solution to the challenge of using powerful indexing languages to reduce noise and conversely increase precision, while limiting execution times to acceptable values. RELIEF takes benefit from the richness of the conceptual graphs formalism [Sow841 and the speed provided by classical information retrieval (IR) techniques. The system is an operational implementation of a logical IR model applied to the conceptual graphs formalism. A sound algebraic approach is proposed that permits to organize the indexes, which am conceptual graphs, into an inverted file-based structure that is then used for faster retrieval [OP98]. The system is integrated on top of the 02 object-oriented DBMS, which provides a complete programming environment, including the fourth generation language OsC, the OQL query language that conforms to the ODMG standard, and the OzLook graphical interface tool. The architecture of our system is shown in figure 1. It supports image handling, by an image manager that makes use of basic 0s classes. The ROGER platform [Oun98] handles conceptual graphs; it is programmed in OsC and an 02Look interface allows for off-line insertion, change, consistency tests and visualization of the indexes. Retrieval can be performed from a Web browser, through our interrogation interface, which uses the OsWeb tool to access the database. To achieve portability and to manage the presentation, an additional JAVA-based layer is provided. The availability of recent advances in the object-oriented database domain and the support of database connections via Web browsers are the main reasons for the choice of 0s OG DBMS. RELIEF is proposed as a solution to the integration of object oriented modeling and Web technologies in the IR domain, based on indexing and interrogation functionalities. l Indexing facilities Image indexing is performed with a computer-assisted pmcess, according to a particular image indexing model [Mec95]. A basic part of the indexing consists in the introduction and

Media objects for user-centered similarity matching

The increase of digital image and video acquisition devices, combined with the growth of the World Wide Web, requires the definition of user-relevant similarity matching methods providing meaningful access to documents searched by users among large amounts of data. The aim of our work is to define media objects for document description suited to images and videos, integrating a user-centered definition of importance for similarity matching. The importance is defined according to criteria and hypotheses, which have been experimentally validated. This leads to a definition of a weighting scheme for media objects (based on objects size, position, and scene homogeneity), which has also been validated with users in a second experiment. This model allows for meaningful similarity matching between document pairs and between users’ queries and documents.