In general, TF-IDF is a most popular method to calculate
relations between terms and documents. Many researches show
that TF-IDF has good effect applied to online library or Q & A
community [1-3]. However, TF-IDF has a disadvantage as
applied to documents, which don’t have enough terms. When
the frequency of important terms is close to common terms,
TF-IDF has a problem that every term appears only one time
easily. That makes difficult to operate the process of a language
model. The title of publications is an obvious example.