E12-1067 multiple split points , we propose a hierarchical segmentation where each segment is split further
J11-1004 The linguistic structure is a hierarchical segmentation of the dialogue . Each segment
D13-1072 The method is based on multiple hierarchical segmentations to sample a limited set of high
N09-1040 efficient inference of the optimal hierarchical segmentation ( given the marginals QZ ) is
N09-1040 . As the goal is to obtain the hierarchical segmentation and not the language models ,
N04-4035 individual stories . While a richer hierarchical segmentation is ultimately desirable , sequential
E03-1058 agglomerative clustering to perform hierarchical segmentation . Another approach to clustering
J97-1005 perform a linear rather than a hierarchical segmentation , where a linear segmentation
N09-1040 a probabilistic setting . For hierarchical segmentation , we take the hypothesis that
D13-1102 new " . We do that by adopting a hierarchical segmentation technique where the same segmentation
J97-1003 theories of discourse assume a hierarchical segmentation model . Foremost among these
N09-1040 to search the entire space of hierarchical segmentations in polynomial time , using a
N09-1040 pyramid is constrained to induce a hierarchical segmentation . Inference takes the form of
N09-1040 baseline , UNIFORM produces a hierarchical segmentation with the ground truth number
D15-1178 ) . We focus on producing rich hierarchical segmentation going beyond clusters which do
N09-1040 topics are constrained to produce a hierarchical segmentation structure , as shown in Figure
D11-1026 modify the algorithm to perform hierarchical segmentation : consider net similarity with
N09-1040 can also be used to extract a hierarchical segmentation , due to the phenomenon of multi-scale
D08-1035 similar Bayesian techniques for hierarchical segmentation , and to incorporate additional
N09-1040 work on dynamic programming for hierarchical segmentation . 4.2 Scale-level marginals The
hide detail