Visual Interfaces for Spatial Databases

We shall apply our background on conceptual modeling and visual interfaces to the development of a prototype for visual interfaces based on the MADS model. The objective is to provide users with a complete visual data definition and data manipulation tool supporting our conceptual paradigm. The tool will automatically map user specifications to at least one and probably several underlying GIS or DBMS. This requires development and/or assessment of:

• translators for MADS data definition into target data definitions (some already exist). Depending on the target system, this may include the automatic generation of methods (or triggers or procedures) to implement MADS specifications which do not have an immediate counterpart in the target system. This is mainly a development task, with limited research issues. Cooperation with GIS providers will be sought.
• a visual schema editor: this work is in progress and should be finalized this year. More sophisticated mechanisms may be added in a later phase (e.g. personnalization of icons, multi-language interface, e.g. English or French or ...). Students projects may contribute with specific additions.
• visual data manipulation facilities using a multi-mode interaction (sketch, text, schema-driven, map-driven, ...). This is first a research task, as no satisfactory visual DML exist by now. We expect that two years will be needed before we can implement a first multi-mode solution, hopefully by joining efforts with other research teams with similar interests (in particular, the query-by-sketch project by Prof. Egenhofer).
• an algebraic DML for spatial databases will be investigated to support the visual DML. Alternatively or complementarily (depending on resources) an OQL-like language may also be developed. The algebraic language is intended to serve as an intermediate language between the visual language and the many target systems (as in the SUPER prototype). Algebraic expressions are an appropriate form for translation to various target languages. The ODL language from the ODMG group may be a priority target. The OQL-like language may be used as an expert user interface.

Further on, models and languages for multiscale databases will be developed, continuing the work which is currently the topic of a PhD in progress. Cooperation with IGN in Paris and the Compagnie Vaudoise d'Electricité will be pursued.

New topics we plan to explore include: modeling of topological integrity constraints, and spatial data mining, which would also enhance the facilities of the visual tools.

We hope to be able to conduct real-life experiments with the visual interface tools, thanks to our cooperation with the Cantonal Administration.

The next step will be to find an industrial partnership for the maketing of a product stemming from our tools. From this viewpoint, we will in particular consider implementing the visual interfaces on top of the GEOTASK product, marketed by the homonym spin-off company. We hope indeed that GEOTASK will serve as a vehicle for smooth transfer of our research results to the marketplace.

Sound archiving and image retrieval

The SIRANAU project seems to be a successful project for all of its partners. We therefore expect the project to continue, despite the fact that the major current source of resources (the funding support by the CTI commission) may dry up by the end of this year.

This is mainly a development activity because of constraints inherent to the project. However, research activity related to the project is being started and at least a PhD is planned on caching issues for sound documents. Another relevant research topic will be sound indexing.

We also hope that sooner or later the project will expand and call for facilities for image archiving and retrieving, thus allowing reuse of the ongoing PhD work on image querying by content. In this case, cooperation with other EPFL labs working on image databases ( LCAV and IMT/ESI) will be sought. Of course, current work on the footprints recognition application will be continued at least up to the end of the PhD in progress, and hopefully beyond in cooperation with users. An analysis and evaluation of various multi-dimensional indexing techniques is also planned as part of this project.

Temporal modeling

The MADS project has shown that spatial applications also need to be able to manage temporal data. Hence, adding the temporal dimension to the MADS model will be the area of major improvement of MADS in the coming years. This work is in progress from the research perspective, and we have established a clear basis for the modeling approach. Topics to be addressed include:

• implementation of the conceptual specifications, in terms of translation to target relational or object-oriented systems
• open issues related to supporting multiple time granularity, including specification of imprecise time management
• open issues related to the management of future time and of branching time
  specification and implementation of a visual DML for temporal data and its underlying algebra.

Conceptual modeling

We still plan to fix some open issues in traditional conceptual modeling, in particular the multi-instantiation aspects. This is becoming a highly appreciated topic, with many contributions from many teams. Still we feel that the approach that we have suggested is a good alternative to existing solutions. The basic idea is to introduce a multi-instantiation link (the may-be link) to complement the is-a link, and to modify the traditional inheritance mechanism (which is currently questioned by many researchers). Interesting results have been achieved within our approach thanks to the PhD theses by Dr Gentile and Dr Ye, but they have not been widely published. An effort should be made to assess the work through major publications.

Another essential issue is the definition of a sound cluster concept for the management of large schemas. This concept is also fundamental to multiscale spatial databases. Most of existing work on this issue is basically diagrammatic, and a good semantic approach is still missing. Some work has been done at the lab as part of a student final work (³travail de diplôme²) or as part of the phD by Dr Auddino. However, the work is incomplete and needs further investigation and implementation.

Database integration for mobile computers

Despite the fact that our research results looked promising, we have not yet been able to start a real implementation of our ideas on database integration. Beyond lack of time, the problem is that this represents a major development work and therefore calls for external funding. Unfortunately, although many enterprises are facing this problem and trying to deal with it on a pure ad-hoc basis, we could not establish a partnership to launch such a project. In the spatial database domain the situation seems more promising, as the cost of data acquisition is so high that the ability to create new data sets through integration of existing disparate ones is of real economical importance. For our potential partners we designed a scenario which states some of the benefits of database integration.

Meanwhile, taking into account the evolving scene of computer usage, in particular the tendency to use wireless portable computers, we believe there is an appropriate setting to deal with database integration issues in a mobile environment. Mobile computers can be used to dynamically connect and disconnect to different databases depending on the actual location of the computer. Every connection raises the problem of understanding what data is available and integrating these different available sources, possibly also with some database on the mobile computer itself. The mobile context thus provides an excellent and most up-to-date environment to test or develop ideas on on-the-fly database integration.

Many research teams are interested in mobile computing, but many are addressing communications issues or connection protocols. We believe there is room for a project on semantic issues like database integration. We shall first have a look on the current relevant literature, and then finalize a proposal for a research project. If the idea materializes, cooperation with other teams (we already have a contact in Australia) will be promoted. The new Institute for Computer Communications and their Applications of our Computer Science Department may also be a perfect partner for such a project.

Databases for mobile objects

Alternatively, our experience on the definition and management of spatio-temporal data may be an incentive to launch a project on databases storing mobile objects, i.e. object whose location changes rapidly and has to be monitored. Application domains for such databases include computer animation and virtual reality systems. Knowing that our Computer Science Department has a very strong group (Prof. Thalmann) in this domain, a project on mobile object DB could foster a solid cooperation between the database lab and Prof. Thalmann's lab. Also, as far as we know there are not so many groups working on this issue. I have visited some of them while in Japan, so some international cooperation is also foreseeable.

Database support for web

We plan to investigate how databases could provide an effective way of developing new web sites based on a patterned approach. A generic pattern for representing the data in a web development application should be investigated first. Next, we plan to explore how information may be dynamically interrelated, grouped, classified. Further focus will be on mechanisms for update propagation and consistency maintenance. Experimentation in real-life environment will be an integral part of this project.