FIX to Faster Compact Top-k Document Retrieval
Level: Medium
Due to a bug in the code, the space and time of our data structure was not
correctly measured. The actual space of our data structure is 3.3-4.0 times
the text size, and the times are about twice those given in the paper. This
puts our structure in a different category, unable to compete with [21] in
space. Rather, it should be compared with plain uncompressed document arrays,
where they use about the same space but our structure is usually 20% to 50%
faster. This makes our structure still an interesting alternative in this
category of "large" representations.