Multistream Dynamic Bayesian Network for Meeting Segmentation

Alfred Dielmann; Steve Renals

doi:10.1007/978-3-540-30568-2_7

Multistream Dynamic Bayesian Network for Meeting Segmentation

Research output: Chapter in Book/Report/Conference proceeding › Chapter

Abstract / Description of output

This paper investigates the automatic analysis and segmentation of meetings. A meeting is analysed in terms of individual behaviours and group interactions, in order to decompose each meeting in a sequence of relevant phases, named meeting actions. Three feature families are extracted from multimodal recordings: prosody from individual lapel microphone signals, speaker activity from microphone array data and lexical features from textual transcripts. A statistical approach is then used to relate low-level features with a set of abstract categories. In order to provide a flexible and powerful framework, we have employed a dynamic Bayesian network based model, characterized by multiple stream processing and flexible state duration modelling. Experimental results demonstrate the strength of this system, providing a meeting action error rate of 9%.

Original language	English
Title of host publication	Machine Learning for Multimodal Interaction
Subtitle of host publication	First International Workshop, MLMI 2004, Martigny, Switzerland, June 21-23, 2004, Revised Selected Papers
Editors	Samy Bengio, Hervé Bourlard
Place of Publication	Berlin, Heidelberg
Publisher	Springer
Pages	76-86
Number of pages	11
ISBN (Electronic)	978-3-540-30568-2
ISBN (Print)	978-3-540-24509-4
DOIs	https://doi.org/10.1007/978-3-540-30568-2_7
Publication status	Published - 2005
Event	First International Workshop MLMI 2004 - Martigny, Switzerland Duration: 21 Jun 2004 → 23 Jun 2004

Publication series

Name	Lecture Notes in Computer Science
Publisher	Springer Berlin Heidelberg
Volume	3361
ISSN (Print)	0302-9743

Workshop

Workshop	First International Workshop MLMI 2004
Country/Territory	Switzerland
City	Martigny
Period	21/06/04 → 23/06/04

Access to Document

10.1007/978-3-540-30568-2_7

dielmann-mlmi04Submitted manuscript, 90 KB

http://dx.doi.org/10.1007/978-3-540-30568-2_7

Cite this

Dielmann, A., & Renals, S. (2005). Multistream Dynamic Bayesian Network for Meeting Segmentation. In S. Bengio, & H. Bourlard (Eds.), Machine Learning for Multimodal Interaction: First International Workshop, MLMI 2004, Martigny, Switzerland, June 21-23, 2004, Revised Selected Papers (pp. 76-86). (Lecture Notes in Computer Science; Vol. 3361). Springer. https://doi.org/10.1007/978-3-540-30568-2_7

Dielmann, Alfred ; Renals, Steve. / Multistream Dynamic Bayesian Network for Meeting Segmentation. Machine Learning for Multimodal Interaction: First International Workshop, MLMI 2004, Martigny, Switzerland, June 21-23, 2004, Revised Selected Papers. editor / Samy Bengio ; Hervé Bourlard. Berlin, Heidelberg : Springer, 2005. pp. 76-86 (Lecture Notes in Computer Science).

@inbook{e8206029a24e452b8a8878ee97863fcf,

title = "Multistream Dynamic Bayesian Network for Meeting Segmentation",

abstract = "This paper investigates the automatic analysis and segmentation of meetings. A meeting is analysed in terms of individual behaviours and group interactions, in order to decompose each meeting in a sequence of relevant phases, named meeting actions. Three feature families are extracted from multimodal recordings: prosody from individual lapel microphone signals, speaker activity from microphone array data and lexical features from textual transcripts. A statistical approach is then used to relate low-level features with a set of abstract categories. In order to provide a flexible and powerful framework, we have employed a dynamic Bayesian network based model, characterized by multiple stream processing and flexible state duration modelling. Experimental results demonstrate the strength of this system, providing a meeting action error rate of 9%.",

author = "Alfred Dielmann and Steve Renals",

year = "2005",

doi = "10.1007/978-3-540-30568-2_7",

language = "English",

isbn = "978-3-540-24509-4",

series = "Lecture Notes in Computer Science",

publisher = "Springer",

pages = "76--86",

editor = "Samy Bengio and Herv{\'e} Bourlard",

booktitle = "Machine Learning for Multimodal Interaction",

address = "United Kingdom",

note = "First International Workshop MLMI 2004 ; Conference date: 21-06-2004 Through 23-06-2004",

}

Dielmann, A & Renals, S 2005, Multistream Dynamic Bayesian Network for Meeting Segmentation. in S Bengio & H Bourlard (eds), Machine Learning for Multimodal Interaction: First International Workshop, MLMI 2004, Martigny, Switzerland, June 21-23, 2004, Revised Selected Papers. Lecture Notes in Computer Science, vol. 3361, Springer, Berlin, Heidelberg, pp. 76-86, First International Workshop MLMI 2004, Martigny, Switzerland, 21/06/04. https://doi.org/10.1007/978-3-540-30568-2_7

Multistream Dynamic Bayesian Network for Meeting Segmentation. / Dielmann, Alfred; Renals, Steve.
Machine Learning for Multimodal Interaction: First International Workshop, MLMI 2004, Martigny, Switzerland, June 21-23, 2004, Revised Selected Papers. ed. / Samy Bengio; Hervé Bourlard. Berlin, Heidelberg: Springer, 2005. p. 76-86 (Lecture Notes in Computer Science; Vol. 3361).

Research output: Chapter in Book/Report/Conference proceeding › Chapter

TY - CHAP

T1 - Multistream Dynamic Bayesian Network for Meeting Segmentation

AU - Dielmann, Alfred

AU - Renals, Steve

PY - 2005

Y1 - 2005

N2 - This paper investigates the automatic analysis and segmentation of meetings. A meeting is analysed in terms of individual behaviours and group interactions, in order to decompose each meeting in a sequence of relevant phases, named meeting actions. Three feature families are extracted from multimodal recordings: prosody from individual lapel microphone signals, speaker activity from microphone array data and lexical features from textual transcripts. A statistical approach is then used to relate low-level features with a set of abstract categories. In order to provide a flexible and powerful framework, we have employed a dynamic Bayesian network based model, characterized by multiple stream processing and flexible state duration modelling. Experimental results demonstrate the strength of this system, providing a meeting action error rate of 9%.

AB - This paper investigates the automatic analysis and segmentation of meetings. A meeting is analysed in terms of individual behaviours and group interactions, in order to decompose each meeting in a sequence of relevant phases, named meeting actions. Three feature families are extracted from multimodal recordings: prosody from individual lapel microphone signals, speaker activity from microphone array data and lexical features from textual transcripts. A statistical approach is then used to relate low-level features with a set of abstract categories. In order to provide a flexible and powerful framework, we have employed a dynamic Bayesian network based model, characterized by multiple stream processing and flexible state duration modelling. Experimental results demonstrate the strength of this system, providing a meeting action error rate of 9%.

U2 - 10.1007/978-3-540-30568-2_7

DO - 10.1007/978-3-540-30568-2_7

M3 - Chapter

SN - 978-3-540-24509-4

T3 - Lecture Notes in Computer Science

SP - 76

EP - 86

BT - Machine Learning for Multimodal Interaction

A2 - Bengio, Samy

A2 - Bourlard, Hervé

PB - Springer

CY - Berlin, Heidelberg

T2 - First International Workshop MLMI 2004

Y2 - 21 June 2004 through 23 June 2004

ER -

Dielmann A, Renals S. Multistream Dynamic Bayesian Network for Meeting Segmentation. In Bengio S, Bourlard H, editors, Machine Learning for Multimodal Interaction: First International Workshop, MLMI 2004, Martigny, Switzerland, June 21-23, 2004, Revised Selected Papers. Berlin, Heidelberg: Springer. 2005. p. 76-86. (Lecture Notes in Computer Science). doi: 10.1007/978-3-540-30568-2_7

Multistream Dynamic Bayesian Network for Meeting Segmentation

Abstract / Description of output

Publication series

Workshop

Access to Document

Fingerprint

Cite this