Combination of classifiers for automatic recognition of dialog acts
Identifieur interne : 005D25 ( Main/Exploration ); précédent : 005D24; suivant : 005D26Combination of classifiers for automatic recognition of dialog acts
Auteurs : Pavel Kral ; Christophe Cerisara ; Jana KleckovaSource :
Abstract
This paper deals with automatic dialog acts (DAs) recognition in Czech. The dialog acts are sentence-level labels that represent different states of a dialogue, depending on the application. Our work focuses on two applications : a multimodal reservation system and an animated talking head for hearing-impaired people. In that context, we consider the following DAs : statements, orders, yes/no questions and other questions. We propose to use both lexical and prosodic information for DAs recognition. The main goal of this paper is to compare different methods to combine the results of both classifiers. On a Czech corpus simulating a reservation of train tickets, the lexical information only gives about 92 % of classification accuracy, while prosody gives only about 45 % of accuracy. When both classifiers are combined with a multilayer perceptron, the lowest (lexical) word error rate further decreases by 26 %. We show that this improvement is close to the optimal one, given the correlation of the lexical and prosodic features. The other combination schemes do not outperform the lexical-only results.
Affiliations:
Links toward previous steps (curation, corpus...)
- to stream Crin, to step Corpus: 004378
- to stream Crin, to step Curation: 004378
- to stream Crin, to step Checkpoint: 000173
- to stream Main, to step Merge: 005F48
- to stream Main, to step Curation: 005D25
Le document en format XML
<record><TEI><teiHeader><fileDesc><titleStmt><title xml:lang="en" wicri:score="577">Combination of classifiers for automatic recognition of dialog acts</title></titleStmt><publicationStmt><idno type="RBID">CRIN:kral05a</idno><date when="2005" year="2005">2005</date><idno type="wicri:Area/Crin/Corpus">004378</idno><idno type="wicri:Area/Crin/Curation">004378</idno><idno type="wicri:explorRef" wicri:stream="Crin" wicri:step="Curation">004378</idno><idno type="wicri:Area/Crin/Checkpoint">000173</idno><idno type="wicri:explorRef" wicri:stream="Crin" wicri:step="Checkpoint">000173</idno><idno type="wicri:Area/Main/Merge">005F48</idno><idno type="wicri:Area/Main/Curation">005D25</idno><idno type="wicri:Area/Main/Exploration">005D25</idno></publicationStmt><sourceDesc><biblStruct><analytic><title xml:lang="en">Combination of classifiers for automatic recognition of dialog acts</title><author><name sortKey="Kral, Pavel" sort="Kral, Pavel" uniqKey="Kral P" first="Pavel" last="Kral">Pavel Kral</name></author><author><name sortKey="Cerisara, Christophe" sort="Cerisara, Christophe" uniqKey="Cerisara C" first="Christophe" last="Cerisara">Christophe Cerisara</name></author><author><name sortKey="Kleckova, Jana" sort="Kleckova, Jana" uniqKey="Kleckova J" first="Jana" last="Kleckova">Jana Kleckova</name></author></analytic></biblStruct></sourceDesc></fileDesc><profileDesc><textClass></textClass></profileDesc></teiHeader><front><div type="abstract" xml:lang="en" wicri:score="3240">This paper deals with automatic dialog acts (DAs) recognition in Czech. The dialog acts are sentence-level labels that represent different states of a dialogue, depending on the application. Our work focuses on two applications : a multimodal reservation system and an animated talking head for hearing-impaired people. In that context, we consider the following DAs : statements, orders, yes/no questions and other questions. We propose to use both lexical and prosodic information for DAs recognition. The main goal of this paper is to compare different methods to combine the results of both classifiers. On a Czech corpus simulating a reservation of train tickets, the lexical information only gives about 92 % of classification accuracy, while prosody gives only about 45 % of accuracy. When both classifiers are combined with a multilayer perceptron, the lowest (lexical) word error rate further decreases by 26 %. We show that this improvement is close to the optimal one, given the correlation of the lexical and prosodic features. The other combination schemes do not outperform the lexical-only results.</div></front></TEI><affiliations><list></list><tree><noCountry><name sortKey="Cerisara, Christophe" sort="Cerisara, Christophe" uniqKey="Cerisara C" first="Christophe" last="Cerisara">Christophe Cerisara</name><name sortKey="Kleckova, Jana" sort="Kleckova, Jana" uniqKey="Kleckova J" first="Jana" last="Kleckova">Jana Kleckova</name><name sortKey="Kral, Pavel" sort="Kral, Pavel" uniqKey="Kral P" first="Pavel" last="Kral">Pavel Kral</name></noCountry></tree></affiliations></record>Pour manipuler ce document sous Unix (Dilib)
EXPLOR_STEP=$WICRI_ROOT/Wicri/Lorraine/explor/InforLorV4/Data/Main/Exploration
HfdSelect -h $EXPLOR_STEP/biblio.hfd -nk 005D25 | SxmlIndent | more
Ou
HfdSelect -h $EXPLOR_AREA/Data/Main/Exploration/biblio.hfd -nk 005D25 | SxmlIndent | more
Pour mettre un lien sur cette page dans le réseau Wicri
{{Explor lien
|wiki= Wicri/Lorraine
|area= InforLorV4
|flux= Main
|étape= Exploration
|type= RBID
|clé= CRIN:kral05a
|texte= Combination of classifiers for automatic recognition of dialog acts
}}
| This area was generated with Dilib version V0.6.33. |  |