One of the biggest concerns of modern information retrieval systems is reducing the user
effort required for manual traversal and filtering of long matching document lists. In this paper
we propose an alternative approach for compact and concise representation of search results,
which we implemented in the BoW on-line bibliographical repository. The BoW repository
is based on an hierarchical concept index to which entries are linked. The key idea is that
searching in the hierarchical repository should take advantage of the repository structure and
return matching topics from the hierarchy, rather than just a long list of entries. Likewise,
when new entries are inserted, a search for relevant topics to which they should be linked is
required. Therefore, a similar hierarchical scheme for query-topic matching can be applied
for both tasks. However, our experiments show that different query types used for these tasks
are best treated by different topic ranking functions. For example, keyword search which
is typically based on short (1-3 word) queries requires a weight-based (rather than Boolean)
ranking approach. The underlying rationale of weight-based ranking is that for a truly relevant
topic all (or almost all) the query terms should appear in its vector representation and with
approximately even high weights. Applying this reasoning to the topic ranking method is
shown to significantly increase the precision and the F1 (by over 30%) for short keyword
queries compared to the baseline Boolean ranking metric.