Detecting and extracting named entities from spontaneous speech in a mixed-initiative spoken dialogue context: How May I Help You? sm;tm Fr  ed eric B echet a, * , Allen L. Gorin b , Jeremy H. Wright b , Dilek Hakkani T€ ur b a LIA, University of Avignon, BP1228, 84911 Avignon Cedex 09, France b AT&T Labs, 180 Park Avenue, Florham Park, NJ 07932, USA Received 17 October 2002; received in revised form 14 June 2003; accepted 1 July 2003 Abstract The understanding module of a spoken dialogue system must extract, from the speech recognizer output, the kind of request expressed by the caller (the call type) and its parameters (numerical expressions, time expressions or proper- names). Such expressions are called Named Entities and their definitions can be either generic or linked to the dialogue application domain. Detecting and extracting such Named Entities within a mixed-initiative dialogue context like How May I Help You? sm;tm (HMIHY) is the subject of this study. After reviewing standard methods based on hand-written grammars and statistical tagging, we propose a new approach, combining the advantages of both in a 2-step process. We also propose a novel architecture which exploits understanding to improve recognition accuracy: the output of the Automatic Speech Recognition module is now a word lattice and the understanding module is responsible for tran- scribing the word strings which are useful to the Dialogue Manager. All the methods proposed are trained and eval- uated on a corpus comprising utterances from live customer traffic. Ó 2003 Elsevier B.V. All rights reserved. R esum e Les syst emes automatiques de dialogue t el  ephonique contiennent g en eralement un module de compr ehension charg e de traiter les sorties du module de reconnaissance automatique de parole. Ce traitement consiste  a extraire non-seu- lement le type de requ^ ete exprim ee par lÕutilisateur mais aussi les param etres de cette requ^ ete tels que les expressions num eriques, temporelles ou bien encore les noms propres. Ces expressions sont g en eralement appele es des Entit es Nomm ees et leurs d efinitions peuvent ^ etre g en eriques ou bien li ees  a un domaine dÕapplication particulier. D etecter et extraire de telles entit esdanslecadredÕun syst eme automatique de dialogue t el  ephonique  a initiative mixte tel que How May I Help You? sm;tm (HMIHY) est le sujet de cette  etude. Apr es avoir pass e en revue les m ethodes habituelles bas ees sur des grammaires  ecrites manuellement ou bien sur des  etiqueteurs statistiques, nous proposons une nouvelle ap- proche permettant de combiner leurs avantages respectifs. Nous proposons  egalement une nouvelle architecture, pour les syst emes automatiques de dialogue t el  ephonique, qui utilize les r esultats du module de compr ehension afin * Corresponding author. Tel.: +33-4-90-84-35-12; fax: +33-4-90-84-35-01. E-mail addresses: frederic.bechet@lia.univ-avignon.fr (F.B echet), algor@research.att.com (A.L. Gorin), jwright@research.att.com (J.H. Wright), dtur@research.att.com (D. Hakkani T€ ur). 0167-6393/$ - see front matter Ó 2003 Elsevier B.V. All rights reserved. doi:10.1016/j.specom.2003.07.003 Speech Communication 42 (2004) 207–225 www.elsevier.com/locate/specom