Supporting Text Retrieval by Typographical Term Weighting - Distributed Artificial Intelligence, Agent Technology, and Collaborative Applications - page 208

Figure 1. Typographical weighting similar to character size

gI guidelines acM guidelines

Title

Title

Section

Subsection
Continuous Text

Section

Subsect on

Continuous Text

Abstract
References

Abstract
References
Footnote

Footnote

Typographical weight Typographical weight

Figure 2. Typographical weighting in levels of used character sizes

gI guidelines

acM guidelines

Title

Title

Section
Subsection
Continuous Text

Section
Subsection
Continuous Text

Abstract
References
Footnote

Abstract
References
Footnote
Typographical weight Typographical weight

by 100 steps. Figure 2 also shows the typographical weighting of the author guidelines of Table 5, but
this time in levels of really used character sizes. Note that this approach assigns similar typographical
weights to comparable sections. Only the missing typographic difference between literature and footnote
sections and the abstract within the ACM guidelines leads to a different weighting.

eVALuATIon

We have evaluated our new typographic term weighting approach by using the widely accepted k near-
est neighbor (kNN) classifier as follows. We combined the classical cosine measure with the matching