Société de gestion d’informations et de documentation

Développement continu de systèmes d’indexation automatisée

Exemples de projets

Développement continu de systèmes d’indexation automatisée

Les systèmes ordinaires d’indexation automatisée ne sont pas assez performants pour extraire des informations très spécifiques de documents scientifiques. La qualité résultante de cette indexation peut être considérablement améliorée si l’on y inclue des informations basées sur une re-évaluation manuelle du contenu. Pour cela GIMD utilise des algorithmes de recherche et de lecture assistée des textes intégraux, permettant à nos employés scientifiques d’identifier rapidement et de façon fiable les informations importantes contenues dans les documents.

Project duration: ongoing

Automatic indexing processes have been used in the most diverse fields for many years. Yet, particularly in the area of LifeSciences, the indexing machines used so far are still not sufficiently mature. The benefits of manual indexing over automated procedures primarily consist in receptive reading, the filtering of relevant information, and, as a result, in greater precision. In addition, manual procedures can also index matters not explicitly named in the respective article or inaccessible to automatic cataloging due to the use of synonyms.

The effectivity of an automatic indexing system may be continually improved, predominately by the adaption of indexing results to high-quality manual indexing. By comparing different indexing results, specific algorithms may be developed, which will increase the precision of the automatic process. The more articles are analyzed and compared, the more finely the specificity of the system may be adjusted.

Automated processes may always involve the problem of generating ballast information, and receptive, in-depth text comprehension does not appear possible at the moment. Still, there are meaningful applications of automatic procedures. By a combination of automatic and manual indexing, we have the long-term hope to be able to offer an even more efficient service to our clients.

In the area of content-based cataloging of scientific literature in Medicine and Pharmacy, we have acquired long-term experience and extremely high competence. We are therefore in a position to test the efficiency of automated processes as quality controllers, to assess them, and to contribute to their further development.