A. Weighting in Documents’ Keyphrase Graph
The representation power of keyphrase graph can be
vastly improved by assigning weights to its keyphrase
vertices. In our previous work, we proposed two weighting
value for document’s keyphrase graph. The “term
frequency” reflect a keyphrase’s importance according to
the number of times it appears in document, and the
“importance of Position” (ip) determines the importance
according to where it appears. However the formula we
chose back then to calculate those weighting frequently
yield a value too small. The small weighting value result in
small similarity evaluation value, thus making search
result ranking harder and may as well impair search
precision.
So after testing and reconsideration, we have revised the
formula for keyphrase weighting. The “term frequency” (tf)
of keyphrase k in the document d will be defined as
follows: