ivan elhart - ece 992 ubiquitous computing university of new hampshire 10/09/2008

22
LEAD REVIEW ON THE PAPER Search Vox: Leveraging Multimodal Refinement and Partial Knowledge for Mobile Voice Search by Tim Peak et al. Ivan Elhart - ECE 992 Ubiquitous Computing University of New Hampshire 10/09/2008

Upload: lavey

Post on 23-Feb-2016

13 views

Category:

Documents


0 download

DESCRIPTION

LEAD REVIEW ON THE PAPER Search Vox : Leveraging Multimodal Refinement and Partial Knowledge for Mobile Voice Search by Tim Peak et al. Ivan Elhart - ECE 992 Ubiquitous Computing University of New Hampshire 10/09/2008. Problems. - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Ivan Elhart - ECE 992 Ubiquitous Computing University of New Hampshire  10/09/2008

LEAD REVIEW ON THE PAPER

Search Vox: Leveraging Multimodal Refinement and

Partial Knowledge for Mobile Voice Search

by

Tim Peak et al.

Ivan Elhart - ECE 992 Ubiquitous ComputingUniversity of New Hampshire 10/09/2008

Page 2: Ivan Elhart - ECE 992 Ubiquitous Computing University of New Hampshire  10/09/2008

Problems Mobile settings often contain non-stationary

noise which cannot be easily canceled Speakers tend to adopt to surrounding noise in

acoustically unhelpful ways

New multimodal interface that will help voice search applications to recover from speech recognition errors

Goal

A multimodal interface for mobile voice search that incorporates speech with touch and text may increase recovery rates of the search

Hypothesis

Page 3: Ivan Elhart - ECE 992 Ubiquitous Computing University of New Hampshire  10/09/2008

Yu, D., Ju, Y.C., Wang, Y.Y., Zweig, G., & Acero, A. 2007. Automated directory assistance system: From theory to practice. Proc. of Interspeech.

Ainsworth, W.A. & Pratt, S.R. 1992. Feedback strategies for error correction in speech recognition systems. International Journal of Man-Machine Studies, 26(6), 833-842.

Hsu, P., Mahajan, M. & Acero, A. 2005. Multimodal text entry on mobile devices. Proc. of ASRU.

Background

Page 4: Ivan Elhart - ECE 992 Ubiquitous Computing University of New Hampshire  10/09/2008

N –best list, whenever recognition is less than perfect

Coupling of speech with touch and text Leveraging of any partial knowledge

Approach

Page 5: Ivan Elhart - ECE 992 Ubiquitous Computing University of New Hampshire  10/09/2008

1. Word palette• Allows users to select any word of a recognized phrase

Approach

Page 6: Ivan Elhart - ECE 992 Ubiquitous Computing University of New Hampshire  10/09/2008

2. Text hints• Resort to speech whenever search entry is too long or

when enough text hints have been provided

Approach

Page 7: Ivan Elhart - ECE 992 Ubiquitous Computing University of New Hampshire  10/09/2008

3. Verbal wildcards• Partial knowledge queries

Approach

Page 8: Ivan Elhart - ECE 992 Ubiquitous Computing University of New Hampshire  10/09/2008

Search Vox architecture

Approach

Page 9: Ivan Elhart - ECE 992 Ubiquitous Computing University of New Hampshire  10/09/2008

Simulation experiments• Utterances collected from Microsoft Live Search Mobile • Automated directory assistance, maps, driving

directions, movie times, local gas prices

Results

Case Frequency

Percentage

Top 1 High Conf (Bull’s Eye)

545 24%

Top 1 Med + Low Conf 1125 48%Top N 183 8%All Wrong 464 20%Total: 2317

Page 10: Ivan Elhart - ECE 992 Ubiquitous Computing University of New Hampshire  10/09/2008

Word palette

Results

Page 11: Ivan Elhart - ECE 992 Ubiquitous Computing University of New Hampshire  10/09/2008

Text hints

Results

Page 12: Ivan Elhart - ECE 992 Ubiquitous Computing University of New Hampshire  10/09/2008

Conclusion

Page 13: Ivan Elhart - ECE 992 Ubiquitous Computing University of New Hampshire  10/09/2008

LEAD REVIEW ON THE PAPER

“It’s Mine, Don’t Touch!”: Interactions at a large Multi-

Touch Display in a City Centre

byPeter Peltonen et al.

Ivan Elhart - ECE 992 Ubiquitous ComputingUniversity of New Hampshire 10/07/2008

Page 14: Ivan Elhart - ECE 992 Ubiquitous Computing University of New Hampshire  10/09/2008

Problem How does the outdoor public tangible interface

support simultaneous participation and interaction of multiple users?

To provide first insights into how users approach, participate, and interact on a large multi-touch display in a public space

Goal

Observational studies in urban environments could help in understanding how multi-touch screens can affect and support social interactions

Public interactive multi-touch displays can potentially restructure the way people experience and use the space around them

Hypotheses

Page 15: Ivan Elhart - ECE 992 Ubiquitous Computing University of New Hampshire  10/09/2008

Semi-public displays

Background

Page 16: Ivan Elhart - ECE 992 Ubiquitous Computing University of New Hampshire  10/09/2008

Public displays

Background

Page 17: Ivan Elhart - ECE 992 Ubiquitous Computing University of New Hampshire  10/09/2008

ApproachCity Wall

Installation in Helsinki

Direct manipulation

Non-modality

Page 18: Ivan Elhart - ECE 992 Ubiquitous Computing University of New Hampshire  10/09/2008

Data collection A continuous interaction log was written Interactions were recorded with a web camera in a

640x480 resolution Twelve on-site interviews were conducted

Data analysis Combination of the video and interaction log First time and returning users Sessions (a ten second gap between interaction)

• Duration• Number of active users• Number of passive bystanders

Approach

Page 19: Ivan Elhart - ECE 992 Ubiquitous Computing University of New Hampshire  10/09/2008

Findings on how the City Wall was used and how the users interacted with each other at the screen

8 days of interaction, 8.8% the display was used, 1199 users, 516 sessions, and 202 passive bystanders

Results

72% 18%23%

Page 20: Ivan Elhart - ECE 992 Ubiquitous Computing University of New Hampshire  10/09/2008

Noticing the Display (when the wall was used)

Multi-user interactions (fun to use with others and friends)

Results

Rain shelter Stepwise app.

Parallel use Team work

Page 21: Ivan Elhart - ECE 992 Ubiquitous Computing University of New Hampshire  10/09/2008

Conflict management

Social configurations (or roles)

Results

Withdrawal Social inter.

Comedian Teacher Leaving a mark

Page 22: Ivan Elhart - ECE 992 Ubiquitous Computing University of New Hampshire  10/09/2008

Conclusion

City Wall

Noticing

Installation

Parallel use Conflicts