Title: Towards a Humanoriented system for Personal Information Management
1Towards a Human-oriented system forPersonal
Information Management
- DPhil Project Audit
- Nuno Jorge Gonçalves de Magalhães Ribeiro
- Supervisor Ian Benest
- Assessor Alistair Edwards
- Department of Computer Science
- University of York
- March 2001
2Motivation (1/2)
- Speech employed in multimedia interfaces for PIM
is potentially useful - when information is visually hidden (e.g.
information lower in hierarchy) - Interactive Guide (interactive multimedia
presentations) - when information is visually disruptive (e.g.
description of a diagram) - Reportage (summarising reports including
multimedia presentations) - when various kinds of reminding and alerting
information must be reported in an integrated and
co-ordinated way w/semantics - Monitor reports notifications (unpredictable
events) - reminders (event alarms with semantics)
3Motivation (2/2)
- Open issues to investigate
- When does speech enhance the user interface and
when is it detrimental to the user interface? - Monitor Reports interrupt
- the users activity
- the audio channel (reportages, interactive guide)
- When to interrupt? How should interruptions be
handled? - Objective to minimise disruption on the user?
4Social Interfaces (Nass and Reeves)
- People react socially to mediated interfaces in
much the same way as they react in face-to-face
situations - If interfaces exhibit supportive modalities that
cue - social responses
- Then people perceive the interaction as more
natural - Humanised Interface
5Aspects of a Humanised Interface
- Interruptions
- (urgency and priority)
- Politeness
- (appropriate level)
- Linguistic Variation
- (user presence and user activity)
-
6Thesis Hypothesis
The incorporation of speech, conveyed using human
discourse characteristics, can improve a
computerised work environment, enabling the
computer to be perceived as a work companion.
7Thesis Hypothesis (1/2)
- 1 Perception of a work companion
- use human discourse characteristics
- linguistic variation
- politeness
- The use of human discourse characteristics such
as linguistic variation and politeness promotes
the perception of a work companion.
8Thesis Hypothesis (2/2)
- 2 Work Environment characteristics
- proper management of interruptions
- interrupt as a human would do
- dont avoid interrupting, but...
- minimise disruption
- interrupt the audio channel
- A proper management of interruptions occurring
in the computer-based work environment promotes
the perception of a work companion.
9Demonstrator design and prototyping
- Design a Speaking Assistant architecture to allow
for - a number of specialised agents (e.g. e-mail,
diary, printer) - a speaking agent
- Speaking agent behaviour
- creates spoken monitor reports (reminders,
notifications) corresponding to messages received - creates reportages (structured multimedia
reports) - interrupts grabs user attention
- fades out on-going reportage
- presents a monitor report
- fades in on-going reportage
10Towards a Personal Assistant in a Multimedia
Environment
Personal Assistant (Specialised Agents)
Speaking Agent infrastructure architecture
Reportage Interactive guide Multimedia clip ...
user activity monitor
context
diary agent
Dispatcher
Aggregator
Vocaliser
MM Engine
User
printer agent
user presence monitor
e-mail agent
Prioritised Messages
Notification Reminder Reportage
Sentences
Spoken messages
Scheduler Priority heuristics
Attention-grabbing Audio channel interruption
Template-based sentence generation
Intonation Prosody generation
11Contributions (1/2)
- How to interrupt the user (activity and audio
channel) - grab the users attention to notify / remind
- fade out, speak, resume at appropriate locations
- must be handled after generating spoken
messages - When to interrupt
- notification mechanism with interruption levels
that depend on urgency, priority and user
activity - must be provided by specialised agents
-
12Contributions (2/2)
- What is an appropriate architecture to support
a multi-agent personal assistant? - a number of specialised agents
- handle specific parts of the user work
environment - send notifications/reminders to be spoken by the
speaking agent - sense the user presence
- a speaking agent
- speaks for a number of specialised agents
- dispatches, aggregates, vocalises and interrupts
- a suitable inter-agent protocol
- what information must specialised agents provide?
- Does this architecture promote the perception
that a system is humanised?
13Evaluation overview
- Assessment of the relative importance of
incorporating some human discourse aspects - at the interface
- Is the system useful? Will people continue to use
it? - Are the generated interruptions appropriate?
- Is the way used to interrupt appropriate?
- Does it appear to be a work companion?
14Evaluation purpose
- What are the beneficial effects of interrupting
using speech at the interface to convey
information that is hidden from view? - What is the role of speech as a contributor to
the human-like nature of the interaction?
15Evaluation empirical study hypotheses (1/3)
- H1 When the system applies an interruption model
based on the urgencies and priorities of
notifications, users will perceive the generated
interruptions as useful (important) and
appropriate (opportune). - Independent variable mode of interruption
interruption model based on urgencies and
priorities / interrupt as events occur. - Dependent variable users perception of the
interruption useful / not useful and
appropriate / not appropriate. - H2 When the system provides spoken messages to
convey notifications and reminders, users feel
less disrupted in their current activity than in
the case where messages are presented graphically
in an independent window.
16Evaluation empirical study hypotheses (2/3)
- H3 Users prefer a fade out/in approach when
the audio-visual channels have to be interrupted
over a full stop of the interrupted audio/video
clips. - H4 Users are less annoyed to be asked about
their availability to receive notifications /
reminders than to have the system interrupting
without regarding their interest in them. - H5 When the system conveys longer and more
detailed messages (with more semantics), users
understand better the purpose of the notification
/ reminder being conveyed through the spoken
message than when spoken messages are shorter or
non-existent (e.g. sound alarms, flashing icons).
17Evaluation empirical study hypotheses (3/3)
- H6 When the system conveys polite messages,
users acceptance is improved and the users
level of irritation is lower than with neutral
messages. - H7 When the system uses linguistic variation,
users acceptance is improved and the users
level of irritation is lower than with repetitive
messages.
18Thesis Outline (1/4)
- Chapter 1 Introduction
- 1.1 Towards a human-oriented system for personal
information management - 1.2 Thesis overview
- Chapter 2 An infrastructure for human-oriented
systems - 2.1 Introduction
- 2.2 A system for personal information management
issues and characteristics - 2.3 The system as a personal assistant
- 2.4 The system as a conversational interface
- 2.5 Summary
19Thesis Outline (2/4)
- Chapter 3 Aspects of a human-oriented system for
personal information management - 3.1 Introduction
- 3.2 Proper management of interruptions
- 3.3 Linguistic variation
- 3.4 Politeness
- 3.5 Summary
- Chapter 4 Design of a conversational assistant
for personal information management - 4.1 Introduction
- 4.2 System overview
- 4.3 A conversational assistant architecture
- 4.4 Speaking agent architecture
- 4.5 Summary
20Thesis Outline (3/4)
- Chapter 5 The conversational assistant prototype
- 5.1 Introduction
- 5.2 Platform
- 5.3 Specialised agents implementation
- 5.4 Messages and the communication protocol
- 5.5 Speaking agent implementation
- 5.6 Using the prototype
- 5.7 Summary
- Chapter 6 Empirical Study
- 6.1 Introduction
- 6.2 Methodology
- 6.3 Results
- 6.4 Discussion
- 6.5 Summary
21Thesis Outline (4/4)
- Chapter 7 Conclusion and future work
- 7.1 Conclusions
- 7.2 Contributions
- 7.3 Future work
- 7.4 Final remarks
- References
- Appendixes
22Work Plan