Jacques Guyot

Automated Patent Classification

Patent classifications are built to set up some order in the growing number and diversity of inventions, and to facilitate patent information searches. The need to automate classification tasks appeared when the growth in the number of patent applications and of classification categories accelerated in a dramatic way. Automated patent classification systems use various elements of patents’ content, which they sort out to find the information most typical of each category. Several algorithms from the field of Artificial Intelligence may be used to perform this task, each of them having its own strengths and weaknesses. Their accuracy is generally evaluated by statistical means. Automated patent classification systems may be used for various purposes, from keeping a classification well organized and consistent, to facilitating some specialized tasks such as prior art search. However, many challenges remain in the years to come to build systems which are more accurate and allow classifying documents in more languages.

Language and Tools to Specify Hypertext Views on Databases

We present a declarative language for the construction of hypertext views on databases. The language is based on an object-oriented data model and a simple hypertext model with reference and inclusion links. A hypertext view specification consists in a collection of parameterized node schemes which specify how to construct node and link instances from the database contents. We show how this language can express different issues in hypertext view design. These include: the direct mapping of objects to nodes; the construction of complex nodes based on sets of objects; the representation of polymorphic sets of objects; and the representation of tree and graph structures. We have defined sublanguages corresponding to particular database models (relational, semantic, object-oriented) and implemented tools to generate Web views for these database models.

Conceptual indexing for multilingual information retrieval

We present a translation-free technique for multilingual information retrieval. This technique is based on an ontological representation of documents and queries. For each language, we use a dictionary (set of lexical reference for concepts) to map a term to its corresponding concept. The same mapping is applied to each document and each query. Then, we use a classic vector space model based on concept for indexing and querying the document corpus. The main advantages of our approach are: no merging phase is required; no dependency on automatic translators between all pairs of languages; and adding a new language only requires a new mapping dictionary to be added into the multilingual ontology. Experimental results on the CLEF 2005 multi8 collection show that this approach is efficient, even with relatively small and low fidelity dictionaries and without word sense disambiguation.

Analysis of word sense disambiguation-based information retrieval

Several studies have tried to improve retrieval performances based on automatic Word Sense Disambiguation techniques. So far, most attempts have failed. We try, through this paper, to give a deep analysis of the reasons behind these failures. During our participation at the Robust WSD task at CLEF 2008, we performed experiments on monolingual (English) and bilingual (Spanish to English) collections. Our official results and a deep analysis are described below, along with our conclusions and perspectives.

A process model for data bases

we present a process model for data bases, which integrates concepts of relation, process, integrity constraint, event and period. The model allows one to specify the synchronization of processes and the constraints due to consumption and production of data. In particular we present the formalization which is carried out with the Petri Nets. These are obtained by using refining primitives which keep the interpretation simple for designers. The aspects of verification and validation of the specification are also examined.

Construire un moteur d'indexation

We present here an indexing engine which is covered by a technology transfer agreement between the University and the private sector. This engine is currently included in various applications used by international organizations. The document collections which are indexed are large and multilingual. The particular elements of the technical specifications are the starting pint of our analysis; then we look at the design and technology choices made to meet the performance and volume constraints. The optimal use of memory, calculations and storage resources is discussed. The serialization and parallelization of processes are analyzed.

Transformation of Urban Knowledge Sources to Ontologies

Since the development of ontologies from scratch requires much time and many resources, the activity of knowledge acquisition constitutes one of the most important steps at the beginning of the ontology development process. This activity is essential in all the different methodologies for ontology design as a previous step to the conceptualization and formalization phases.

TOK: A Meta-model and Ontology for Heterogeneous Terminological, Linguistic and Ontological Knowledge Resources

Documents are rich resources containing knowledge describing a specific domain. Thats why their processing is a common task, which is based on the use of terminological and ontological resources. Various types of ontologies, thesauri, and a large list of resources are commonly used in the process of knowledge extraction. The modeling and reuse of these resources is intended to support knowledge management. In this paper, we propose a methodology and a model for ontological and terminological resource management. Our aim is to build a resources repository that offers operations for loading, storing, indexing, translating, generating and matching different resources. In this contribution we propose an ontology as a model of these resources and we explain how can we represent, annotate and load new resources into our repository.

Ontologies and Multilingualism

The definition of an ontology as a specification of a conceptualization of a domain is independent of the terminology used in a particular natural language to describe this domain. In fact we can make a clear distinction between the conceptual structure of a domain and the way the concepts are designated by terms in a natural language. This view is exemplified in ontology specification languages such as OWL in which there is no connection with terms or texts in natural language, except for comments. In such a language, an ontology designer can arbitrarily define new concepts that do not correspond to any term in an existing language.

UniGE experiments on prior art search in the field of patents

In this experiment led at the University of Geneva (UniGE), we evaluated several similarity measures as well as the relevance of using automated classification to filter out search results. The patent field is particularly well suited to classification-based filtering because each patent is already classified. Our results show that such a filtering approach does not improve searching performances, but it does not have a negative impact on recall either. This last observation allows considering classification as a possible tool to reduce the search space without reducing the quality of search results.