your are in: home - classification and indexing

Press Releases

xinexus develops nuclear archive for IAEA

Download brochures

Contact us

tel: +41 (0)71 929 77 44

email: info@xinexus.com

How to find us

Impressum

Disclaimer

Career opportunities

xinexus NUCLEAR AG combines domain knowledge in the nuclear field with unique expertise in advanced technology for indexing and document classification.

Indexing and Classification of Information

Precise document classification and content indexing add considerable power and precision to knowledge retrieval. But is it affordable? Detailed content analysis by subject specialists is expensive and often carries a bias towards the personal style of the human indexer.

Computer systems on the other hand, used to lack the sophistication to deal with information in context. In recent years, however, this situation has changed: computer assisted indexing, fully automated indexing and dynamic document classification are now used by many key players in the information and knowledge domain.

Keyword indexing

The value of keyword indexing, compared to full-text indexing, lies in the precision of a controlled vocabulary. Experts who index documents in their own field of work with "free" keywords, are implicitly assumed to utilize a professional vocabulary where terms have specific meaning that can be quite different from the every day use of those terms. Even better results can be achieved, if the keywords are taken from a thesaurus which not only restricts the range of vocabulary, but also uniquely defines the meaning of each term through term relations and scope notes.

In the past, automated indexers simply looked for matches between terms found in a piece of text and the controlled vocabulary, possibly taking stemming or grammatical rules into account. The results of such primitive indexing are accordingly poor.

xinexus NUCLEAR AG works with fully semantic-sensitive indexing tools that recognise not only word variations (spelling, grammar, phonetic matching) but also consider each term's semantic context both in the thesaurus as well as in the document.

When INIS (http://www.iaea.org/inis/), the International Nuclear Information System at the IAEA recently commissioned its Computer Aided Indexing System, CAI, it was found that CAI indexed not only much more comprehensively than human indexers do, but that the relevance of the keywords, too, was between 90% and 100%. Although INIS, being the largest and most reputed nuclear information system in the world, still relies on human experts as the final instance in quality control, the productivity of its subject specialists increased several fold since the implementation of the new system. Fully automatic indexing, without loss of quality and entailing a significant cost saving combined with increased productivity, has now become a realistic option for INIS.

xinexus NUCLEAR AG supplies and fully customises not only the indexing tools for you, we also support you with the thesaurus creation and maintenance, its multi-lingual implementation and with the creation of special term-variant enumerations for domain specific concept recognition.

Dynamic classification

The fully automatic classification of documents by computer systems builds on concepts that are used for automated keyword indexing. Instead of thesauri, strictly hierarchical taxonomies are used as the basic semantic building blocks. Taxonomies are again tagged with lists of keywords and they can be combined to create more complex classification systems.

The end-user of the system will be able to take advantage of the automatic classification in two ways. The classification can be used to restrict the document base in order to increase the precision of search queries, or search results can be used to automatically populate a browsable structure of returned documents. The graphic below illustrates an application, where search results are automatically represented in a table structure, using partial taxonomies for reactors by location and for accidents.

The user can 'walk' through the two taxonomies by clicking on the names of the respective nodes on the horizontal and vertical axis, and he or she can eventually display specific search results by clicking on the document symbol in the matrix elements of the table.

For further information, clease contact us at info@xinexus.com.