PlanetP's search efficacy (contd.)
This results were omitted from the
paper because of the lack of space. They show the average recall
and precision obtained for each of the document collections
studied. In each graph we compare against a simulated centralized
search engine that uses TFxIDF and we also show how the document
distribution affects these two metriecs. The legends X.Y.Z are decoded as follows: X =
{T: search engine using TFxIDF, P: PlanetP}, Y ={W: Weibull, U:
Uniform}, and Z = {z: indexed the most frequently appearing z% of the unique terms in each
document}; for example, T.W.100 means TFxIDF running on a Weibull distribution of documents, where all 100\%
of the unique terms of each document was indexed. |