... expectation and itsmaximization are easy to compute. This is becauseour dialog model has a chain-like structure thatclosely resembles an Hidden Markov Model, so a forward-backward procedure can ... 08540, USAusyed@cs.princeton.eduJason D. WilliamsShannon LaboratoryAT&T Labs — ResearchFlorham Park, NJ 07932, USAjdw@research.att.comAbstractWe use an EM algorithm to learn user ... θ be the maximum likelihoodestimate using automatically transcribed data,i.e., θasu=eKDasP a eKDas. This approach ignorestranscription errors and assumes that user be-havior depends...
... Knowledge BaseYoji Kiyota, Sadao Kurohashi (The University of Tokyo)kiyota,kuro @kc.t.u-tokyo.ac.jpTeruhisa Misu, Kazunori Komatani, Tatsuya Kawahara (Kyoto University)misu,komatani,kawahara @kuis.kyoto-u.ac.jpFuyuko ... typical vague user question. If a userquestion matches it, the dialog manager asks theback question afterSYS , showing choices be- Dialog Navigator : ASpokenDialog Q -A System based on Large ... @kuis.kyoto-u.ac.jpFuyuko Kido (Microsoft Co., Ltd.)fkido@microsoft.comAbstractThis paper describes aspokendialog Q- A system as a substitution for call centers.The system is capable of making dialogsfor...
... M.Walker, D.J.Litman, C .A. Kamm, and A. Abella. 1997. PARADISE: A framework for evaluating spoken dialogue agents. Tech- nical Report TR 97.26.1, AT and T Technical Reports. D.Gates, A. Lavie, ... individual system components separately (Simpson and Fraser, 1993). 571 Designing a Task-Based Evaluation Methodology for aSpoken Machine Translation System Kavita Thomas Language Technologies ... that we can separate in-domain goals from cross-domain goals; cross-domain goals of- ten serve a meta-level purpose in the dialogue. We can thus evaluate performance over all goals while maintaining...
... a human-human dialog leading to more natural dialogs with themachine.4.7 Integration into the ontologyThe foundational ontology (Cimiano et al., 2004)integrated into the dialogsystem Smartweb ... ontology of the system, but to make the dialog more natural andtherefore user-friendly.Natural language utterances processed by anopen-domain spokendialogsystem may containwords or parts of words ... task-oriented evaluation setup and toameliorate the results with various techniques.As natural language texts are not only rich in hi-erarchical relations but in other semantic relationsas...
... theoptions are American, Pizza, and Italian.Most of them are located in Boston andCambridge.SIM: Any restaurants in Back Bay?SYS: There are 57 restaurants in Back Bay.Many of them are American, and ... yields greater robustness against changes in thedatabase contents.3.2 Dialog ManagementThe domain-independent dialog manager is config-urable via an external dialog control table. A setof ... Ferguson and Allen, 1998), and systemsthat enable querying across large and frequentlychanging databases.The goal of this work is to assemble natural spo-ken dialog interfaces that enable flexible...
... INTERSPEECH, pages 268–271.C. Hori, K. Ohtake, T. Misu, H. Kashioka, and S. Naka-mura. 2009b. Statistical dialog management appliedto wfst-based dialog systems. In Proc. IEEE Inter-national Conference ... under-standing (SLU) module converts Wtto an abstractrepresentation of the user’s dialogue act (DA). Thedialogue management (DM) module determines theuser’s dialogue act A ∗tand accordingly ... 1991. Simulating speechsystems. Computer Speech and Language, 5(1):81–99.C. Hori, K. Ohtake, T. Misu, H. Kashioka, and S. Naka-mura. 200 9a. Recent advances in wfst-based dialog system. In...
... each paid reward.• Qualifications To improve the data quality, a HIT can also be attached to certain tests,“qualifications” that are either system- providedor created by the requester. An example ... excess of information. FAQ-pages tend to alsoanswer questions which are not asked, and also con-tain practical examples. Human-powered answersoften contain unrelated information and discourse-like ... the assign-ments have been completed.• Rewards At upload time, each HIT has to beassigned a fixed reward, that cannot be changedlater. Minimum reward is $0.01. Amazon.comcollects a 10% (or a...
... anapproach to spoken dialogue systems thatincludes rule-based and trainable dialoguemanagers, spoken language understandingand generation modules, and a compre-hensive dialogue system architecture. ... information at various levels of descrip-tion in a uniform way. This facilitates dialog eval-uation, data mining and online learning becausedata is available for querying as soon as it hasbeen ... Simulations andRule-based Dialogue Management in aSpoken Dialogue System Sebastian Varges, Silvia Quarteroni, Giuseppe Riccardi, Alexei V. Ivanov, Pierluigi RobertiDepartment of Information...
... restaurants.Above has good decor, and Carmine’shas decent decor. Above and Carmine’shave good service. Above is a NewAmerican restaurant. On the otherhand, Carmine’s is an Italian restau-rant.2.5 ... knowledgeautomatically adapted to the application do-main. This paper presents a trainable sentenceplanner for the MATCH dialog system. Weshow that trainable sentence planning can pro-duce ... Carmine’s andAbove have good service. Carmine’s isan Italian restaurant. Above, however,is a New American restaurant.2.5 .4925 Above and Carmine’s offer exceptionalvalue among the selected restaurants.Above...
... 02-01. Availablefrom http://ilk.kub.nl/downloads/pub/papers/ilk0201.ps.gz.Staffan Larsson and Stina Ericsson. 2002. GoDiS– Issue-Based Dialogue Management in a Multi-Domain, Multi-Language Dialogue ... Las Palmas,Gran Canaria.Walter Daelemans, Jakub Zavrel, Ko van der Sloot,and Antal van den Bosch. 2002. TIMBL: TilburgMemmory Based Learner, version 4.2, ReferenceGuide. In ILK Technical ... a false neg-ative to the cost of a false positive. This is exactlywhat we want but we found that we cannot linearlyvary this parameter in a way that gives us a smoothtransition between false...
... separate data area is reserved for this purpose.The separate data area of these database systems meansthat they do not need the segment cleaning mechanisms ofthe Sprite LFS to reclaim log space. ... under grant CCR-8900029, and in partby the National Aeronautics and Space Administration and theDefense Advanced Research Projects Agency under contractNAG2-591.This paper will appear in the ... illustrates the fact that a log-structured file system produces a different form of locality on disk thantraditional file systems. A traditional file system achieveslogical locality by assuming certain...