Related Experiment Videos
SOPHIA: an interactive cluster-based retrieval system for the OHSUMED collection.
Vladimir Dobrynin1, David Patterson, Mykola Galushka
1Department of Programming Technology, Saint Petersburg State University, St. Petersburg, Russia. vdobr@oasis.apmath.spbu.ru
Summary
We developed SOPHIA, an unsupervised clustering technique for efficient document retrieval. This method visually clusters large document sets, improving search in specialized fields like medicine.
Area of Science:
- Information Science
- Computer Science
- Medical Informatics
Background:
- Effective exploratory search and retrieval of relevant documents from large, domain-specific collections are crucial in medicine and other fields.
- Existing methods may lack semantic meaningfulness or intuitive interaction for complex document sets.
Purpose of the Study:
- To introduce SOPHIA, an unsupervised distributional clustering technique.
- To provide semantically meaningful visual clustering of document corpora.
- To enable intuitive interactive search facilities for enhanced information retrieval.
Main Methods:
- SOPHIA employs unsupervised distributional clustering.
- The technique generates visual clusters of the document corpus.
- An interactive search facility is integrated with the clustering.
Main Results:
- SOPHIA offers semantically meaningful visual clustering.
- The system provides an intuitive interactive search experience.
- Effectiveness was assessed using the OHSUMED test collection for MEDLINE document retrieval.
Conclusions:
- SOPHIA demonstrates a viable approach for cluster-based information retrieval.
- The technique enhances the exploratory search of large medical document collections.
- Visual and interactive features improve the usability of document retrieval systems.