...corpus.

DF frequencies in Switchboard were estimated from a hand-labeled subset of 60 conversation sides, containing 40,500 words. The coverage figure takes into account the further limits on modeled repetitions and utterance-medial deletions described below.

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

...segments.

 A preliminary version of annotated Switchboard data was made available to the 1995 Johns Hopkins Language Modeling Workshop; the LDC will release a final version.

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

...models.

Both baseline and DF models were trained on the same data, which corresponds to only a portion of the full training corpus. Therefore, the perplexity figures are higher here than in some of the comparisons below.

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

...data.

Due to differences in amount of training data and type of segmentation, the perplexities are not directly comparable to the previous two studies.

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