In this paper , we investigate the problem of automatically predicting
<term>
segment boundaries
</term>
in
<term>
spoken multiparty dialogue
</term>
.
#11410In this paper, we investigate the problem of automatically predictingsegment boundaries in spoken multiparty dialogue.
other,14-1-E06-1035,ak
In this paper , we investigate the problem of automatically predicting
<term>
segment boundaries
</term>
in
<term>
spoken multiparty dialogue
</term>
.
#11413In this paper, we investigate the problem of automatically predicting segment boundaries inspoken multiparty dialogue.
other,10-3-E06-1035,ak
We first apply approaches that have been proposed for predicting
<term>
top-level topic shifts
</term>
to the problem of identifying
<term>
subtopic boundaries
</term>
.
#11435We first apply approaches that have been proposed for predictingtop-level topic shifts to the problem of identifying subtopic boundaries.
other,18-3-E06-1035,ak
We first apply approaches that have been proposed for predicting
<term>
top-level topic shifts
</term>
to the problem of identifying
<term>
subtopic boundaries
</term>
.
#11443We first apply approaches that have been proposed for predicting top-level topic shifts to the problem of identifyingsubtopic boundaries.
other,9-4-E06-1035,ak
We then explore the impact on performance of using
<term>
ASR output
</term>
as opposed to
<term>
human transcription
</term>
.
#11455We then explore the impact on performance of usingASR output as opposed to human transcription.
other,5-5-E06-1035,ak
Examination of the effect of
<term>
features
</term>
shows that predicting top-level and predicting
<term>
subtopic boundaries
</term>
are two distinct tasks : ( 1 ) for predicting
<term>
subtopic boundaries
</term>
, the
<term>
lexical cohesion-based approach
</term>
alone can achieve competitive results , ( 2 ) for predicting
<term>
top-level boundaries
</term>
, the
<term>
machine learning approach
</term>
that combines
<term>
lexical-cohesion and conversational features
</term>
performs best , and ( 3 )
<term>
conversational cues
</term>
, such as
<term>
cue phrases
</term>
and
<term>
overlapping speech
</term>
, are better indicators for the
<term>
top-level prediction task
</term>
.
#11468Examination of the effect offeatures shows that predicting top-level and predicting subtopic boundaries are two distinct tasks: (1) for predicting subtopic boundaries, the lexical cohesion-based approach alone can achieve competitive results, (2) for predicting top-level boundaries, the machine learning approach that combines lexical-cohesion and conversational features performs best, and (3) conversational cues, such as cue phrases and overlapping speech, are better indicators for the top-level prediction task.
other,12-5-E06-1035,ak
Examination of the effect of
<term>
features
</term>
shows that predicting top-level and predicting
<term>
subtopic boundaries
</term>
are two distinct tasks : ( 1 ) for predicting
<term>
subtopic boundaries
</term>
, the
<term>
lexical cohesion-based approach
</term>
alone can achieve competitive results , ( 2 ) for predicting
<term>
top-level boundaries
</term>
, the
<term>
machine learning approach
</term>
that combines
<term>
lexical-cohesion and conversational features
</term>
performs best , and ( 3 )
<term>
conversational cues
</term>
, such as
<term>
cue phrases
</term>
and
<term>
overlapping speech
</term>
, are better indicators for the
<term>
top-level prediction task
</term>
.
#11475Examination of the effect of features shows that predicting top-level and predictingsubtopic boundaries are two distinct tasks: (1) for predicting subtopic boundaries, the lexical cohesion-based approach alone can achieve competitive results, (2) for predicting top-level boundaries, the machine learning approach that combines lexical-cohesion and conversational features performs best, and (3) conversational cues, such as cue phrases and overlapping speech, are better indicators for the top-level prediction task.
other,24-5-E06-1035,ak
Examination of the effect of
<term>
features
</term>
shows that predicting top-level and predicting
<term>
subtopic boundaries
</term>
are two distinct tasks : ( 1 ) for predicting
<term>
subtopic boundaries
</term>
, the
<term>
lexical cohesion-based approach
</term>
alone can achieve competitive results , ( 2 ) for predicting
<term>
top-level boundaries
</term>
, the
<term>
machine learning approach
</term>
that combines
<term>
lexical-cohesion and conversational features
</term>
performs best , and ( 3 )
<term>
conversational cues
</term>
, such as
<term>
cue phrases
</term>
and
<term>
overlapping speech
</term>
, are better indicators for the
<term>
top-level prediction task
</term>
.
#11487Examination of the effect of features shows that predicting top-level and predicting subtopic boundaries are two distinct tasks: (1) for predictingsubtopic boundaries, the lexical cohesion-based approach alone can achieve competitive results, (2) for predicting top-level boundaries, the machine learning approach that combines lexical-cohesion and conversational features performs best, and (3) conversational cues, such as cue phrases and overlapping speech, are better indicators for the top-level prediction task.
tech,28-5-E06-1035,ak
Examination of the effect of
<term>
features
</term>
shows that predicting top-level and predicting
<term>
subtopic boundaries
</term>
are two distinct tasks : ( 1 ) for predicting
<term>
subtopic boundaries
</term>
, the
<term>
lexical cohesion-based approach
</term>
alone can achieve competitive results , ( 2 ) for predicting
<term>
top-level boundaries
</term>
, the
<term>
machine learning approach
</term>
that combines
<term>
lexical-cohesion and conversational features
</term>
performs best , and ( 3 )
<term>
conversational cues
</term>
, such as
<term>
cue phrases
</term>
and
<term>
overlapping speech
</term>
, are better indicators for the
<term>
top-level prediction task
</term>
.
#11491Examination of the effect of features shows that predicting top-level and predicting subtopic boundaries are two distinct tasks: (1) for predicting subtopic boundaries, thelexical cohesion-based approach alone can achieve competitive results, (2) for predicting top-level boundaries, the machine learning approach that combines lexical-cohesion and conversational features performs best, and (3) conversational cues, such as cue phrases and overlapping speech, are better indicators for the top-level prediction task.
other,42-5-E06-1035,ak
Examination of the effect of
<term>
features
</term>
shows that predicting top-level and predicting
<term>
subtopic boundaries
</term>
are two distinct tasks : ( 1 ) for predicting
<term>
subtopic boundaries
</term>
, the
<term>
lexical cohesion-based approach
</term>
alone can achieve competitive results , ( 2 ) for predicting
<term>
top-level boundaries
</term>
, the
<term>
machine learning approach
</term>
that combines
<term>
lexical-cohesion and conversational features
</term>
performs best , and ( 3 )
<term>
conversational cues
</term>
, such as
<term>
cue phrases
</term>
and
<term>
overlapping speech
</term>
, are better indicators for the
<term>
top-level prediction task
</term>
.
#11505Examination of the effect of features shows that predicting top-level and predicting subtopic boundaries are two distinct tasks: (1) for predicting subtopic boundaries, the lexical cohesion-based approach alone can achieve competitive results, (2) for predictingtop-level boundaries, the machine learning approach that combines lexical-cohesion and conversational features performs best, and (3) conversational cues, such as cue phrases and overlapping speech, are better indicators for the top-level prediction task.
tech,46-5-E06-1035,ak
Examination of the effect of
<term>
features
</term>
shows that predicting top-level and predicting
<term>
subtopic boundaries
</term>
are two distinct tasks : ( 1 ) for predicting
<term>
subtopic boundaries
</term>
, the
<term>
lexical cohesion-based approach
</term>
alone can achieve competitive results , ( 2 ) for predicting
<term>
top-level boundaries
</term>
, the
<term>
machine learning approach
</term>
that combines
<term>
lexical-cohesion and conversational features
</term>
performs best , and ( 3 )
<term>
conversational cues
</term>
, such as
<term>
cue phrases
</term>
and
<term>
overlapping speech
</term>
, are better indicators for the
<term>
top-level prediction task
</term>
.
#11509Examination of the effect of features shows that predicting top-level and predicting subtopic boundaries are two distinct tasks: (1) for predicting subtopic boundaries, the lexical cohesion-based approach alone can achieve competitive results, (2) for predicting top-level boundaries, themachine learning approach that combines lexical-cohesion and conversational features performs best, and (3) conversational cues, such as cue phrases and overlapping speech, are better indicators for the top-level prediction task.
other,51-5-E06-1035,ak
Examination of the effect of
<term>
features
</term>
shows that predicting top-level and predicting
<term>
subtopic boundaries
</term>
are two distinct tasks : ( 1 ) for predicting
<term>
subtopic boundaries
</term>
, the
<term>
lexical cohesion-based approach
</term>
alone can achieve competitive results , ( 2 ) for predicting
<term>
top-level boundaries
</term>
, the
<term>
machine learning approach
</term>
that combines
<term>
lexical-cohesion and conversational features
</term>
performs best , and ( 3 )
<term>
conversational cues
</term>
, such as
<term>
cue phrases
</term>
and
<term>
overlapping speech
</term>
, are better indicators for the
<term>
top-level prediction task
</term>
.
#11514Examination of the effect of features shows that predicting top-level and predicting subtopic boundaries are two distinct tasks: (1) for predicting subtopic boundaries, the lexical cohesion-based approach alone can achieve competitive results, (2) for predicting top-level boundaries, the machine learning approach that combineslexical-cohesion and conversational features performs best, and (3) conversational cues, such as cue phrases and overlapping speech, are better indicators for the top-level prediction task.
other,62-5-E06-1035,ak
Examination of the effect of
<term>
features
</term>
shows that predicting top-level and predicting
<term>
subtopic boundaries
</term>
are two distinct tasks : ( 1 ) for predicting
<term>
subtopic boundaries
</term>
, the
<term>
lexical cohesion-based approach
</term>
alone can achieve competitive results , ( 2 ) for predicting
<term>
top-level boundaries
</term>
, the
<term>
machine learning approach
</term>
that combines
<term>
lexical-cohesion and conversational features
</term>
performs best , and ( 3 )
<term>
conversational cues
</term>
, such as
<term>
cue phrases
</term>
and
<term>
overlapping speech
</term>
, are better indicators for the
<term>
top-level prediction task
</term>
.
#11525Examination of the effect of features shows that predicting top-level and predicting subtopic boundaries are two distinct tasks: (1) for predicting subtopic boundaries, the lexical cohesion-based approach alone can achieve competitive results, (2) for predicting top-level boundaries, the machine learning approach that combines lexical-cohesion and conversational features performs best, and (3)conversational cues, such as cue phrases and overlapping speech, are better indicators for the top-level prediction task.
other,67-5-E06-1035,ak
Examination of the effect of
<term>
features
</term>
shows that predicting top-level and predicting
<term>
subtopic boundaries
</term>
are two distinct tasks : ( 1 ) for predicting
<term>
subtopic boundaries
</term>
, the
<term>
lexical cohesion-based approach
</term>
alone can achieve competitive results , ( 2 ) for predicting
<term>
top-level boundaries
</term>
, the
<term>
machine learning approach
</term>
that combines
<term>
lexical-cohesion and conversational features
</term>
performs best , and ( 3 )
<term>
conversational cues
</term>
, such as
<term>
cue phrases
</term>
and
<term>
overlapping speech
</term>
, are better indicators for the
<term>
top-level prediction task
</term>
.
#11530Examination of the effect of features shows that predicting top-level and predicting subtopic boundaries are two distinct tasks: (1) for predicting subtopic boundaries, the lexical cohesion-based approach alone can achieve competitive results, (2) for predicting top-level boundaries, the machine learning approach that combines lexical-cohesion and conversational features performs best, and (3) conversational cues, such ascue phrases and overlapping speech, are better indicators for the top-level prediction task.
other,70-5-E06-1035,ak
Examination of the effect of
<term>
features
</term>
shows that predicting top-level and predicting
<term>
subtopic boundaries
</term>
are two distinct tasks : ( 1 ) for predicting
<term>
subtopic boundaries
</term>
, the
<term>
lexical cohesion-based approach
</term>
alone can achieve competitive results , ( 2 ) for predicting
<term>
top-level boundaries
</term>
, the
<term>
machine learning approach
</term>
that combines
<term>
lexical-cohesion and conversational features
</term>
performs best , and ( 3 )
<term>
conversational cues
</term>
, such as
<term>
cue phrases
</term>
and
<term>
overlapping speech
</term>
, are better indicators for the
<term>
top-level prediction task
</term>
.
#11533Examination of the effect of features shows that predicting top-level and predicting subtopic boundaries are two distinct tasks: (1) for predicting subtopic boundaries, the lexical cohesion-based approach alone can achieve competitive results, (2) for predicting top-level boundaries, the machine learning approach that combines lexical-cohesion and conversational features performs best, and (3) conversational cues, such as cue phrases andoverlapping speech, are better indicators for the top-level prediction task.
other,78-5-E06-1035,ak
Examination of the effect of
<term>
features
</term>
shows that predicting top-level and predicting
<term>
subtopic boundaries
</term>
are two distinct tasks : ( 1 ) for predicting
<term>
subtopic boundaries
</term>
, the
<term>
lexical cohesion-based approach
</term>
alone can achieve competitive results , ( 2 ) for predicting
<term>
top-level boundaries
</term>
, the
<term>
machine learning approach
</term>
that combines
<term>
lexical-cohesion and conversational features
</term>
performs best , and ( 3 )
<term>
conversational cues
</term>
, such as
<term>
cue phrases
</term>
and
<term>
overlapping speech
</term>
, are better indicators for the
<term>
top-level prediction task
</term>
.
#11541Examination of the effect of features shows that predicting top-level and predicting subtopic boundaries are two distinct tasks: (1) for predicting subtopic boundaries, the lexical cohesion-based approach alone can achieve competitive results, (2) for predicting top-level boundaries, the machine learning approach that combines lexical-cohesion and conversational features performs best, and (3) conversational cues, such as cue phrases and overlapping speech, are better indicators for thetop-level prediction task.
other,5-6-E06-1035,ak
We also find that the
<term>
transcription errors
</term>
inevitable in
<term>
ASR output
</term>
have a negative impact on
<term>
models
</term>
that combine
<term>
lexical-cohesion and conversational features
</term>
, but do not change the general preference of approach for the two tasks .
#11550We also find that thetranscription errors inevitable in ASR output have a negative impact on models that combine lexical-cohesion and conversational features, but do not change the general preference of approach for the two tasks.
other,9-6-E06-1035,ak
We also find that the
<term>
transcription errors
</term>
inevitable in
<term>
ASR output
</term>
have a negative impact on
<term>
models
</term>
that combine
<term>
lexical-cohesion and conversational features
</term>
, but do not change the general preference of approach for the two tasks .
#11554We also find that the transcription errors inevitable inASR output have a negative impact on models that combine lexical-cohesion and conversational features, but do not change the general preference of approach for the two tasks.
model,16-6-E06-1035,ak
We also find that the
<term>
transcription errors
</term>
inevitable in
<term>
ASR output
</term>
have a negative impact on
<term>
models
</term>
that combine
<term>
lexical-cohesion and conversational features
</term>
, but do not change the general preference of approach for the two tasks .
#11561We also find that the transcription errors inevitable in ASR output have a negative impact onmodels that combine lexical-cohesion and conversational features, but do not change the general preference of approach for the two tasks.
other,19-6-E06-1035,ak
We also find that the
<term>
transcription errors
</term>
inevitable in
<term>
ASR output
</term>
have a negative impact on
<term>
models
</term>
that combine
<term>
lexical-cohesion and conversational features
</term>
, but do not change the general preference of approach for the two tasks .
#11564We also find that the transcription errors inevitable in ASR output have a negative impact on models that combinelexical-cohesion and conversational features, but do not change the general preference of approach for the two tasks.