... 2''^2.1

``based on average 5-point $F_{1}$ score averaged across compression levels and normalized with the random sentence-selection baseline.'' [13]

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... evaluation.^3.1

Hearst notes, however, that there is a high level of disagreement between human judges in a topic segmentation task, and thus this method is not infallible [6]. See Chapter 5 for a discussion of this issue.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.