Top Banner
Stefan Marti Speech Interface Group MIT Media Lab
21

Stefan Marti Speech Interface Group MIT Media Lab.

Jan 05, 2016

Download

Documents

Ella Ramsey
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Page 1: Stefan Marti Speech Interface Group MIT Media Lab.

Stefan Marti

Speech Interface Group

MIT Media Lab

Page 2: Stefan Marti Speech Interface Group MIT Media Lab.

What We All WantWhat We All Want

The ultimate communication device!

Universal, has everything built inWorks everywhere, anytimeLong range, never runs out of battery,

etc

Page 3: Stefan Marti Speech Interface Group MIT Media Lab.

Solution?Solution?

The CommBadge!

FOR MORE INFO...

Please watch the TV show Star Trek.

Above model will be available at around 2364…

Page 4: Stefan Marti Speech Interface Group MIT Media Lab.

Instead: Current SolutionInstead: Current Solution

Cellphones, pagers, other cellphones, more pagers, wireless PDAs, etc.

Page 5: Stefan Marti Speech Interface Group MIT Media Lab.

Actually, there’s even more…Actually, there’s even more…

Fax machines, wired phones, desktop computers, etc.

Page 6: Stefan Marti Speech Interface Group MIT Media Lab.

Why this variety?Why this variety?

We don’t have the technology to pack everything into one device (will change in the future)

Smallest devices will never be universal (wrist, finger ring, ear ring)

AND: No one device fits all situations!

Page 7: Stefan Marti Speech Interface Group MIT Media Lab.

Our Solution!Our Solution!

Something that integrates all our existing communication channels and devices:

Page 8: Stefan Marti Speech Interface Group MIT Media Lab.

Active Messenger (AM) is a kind of agent software that deals with a user’s incoming email

AM manages this user’s communication channels

AM is used by two users for two years now

Page 9: Stefan Marti Speech Interface Group MIT Media Lab.

AM guarantees the delivery of messages

AM modifies its filtering and forwarding rules depending on which devices are being used and current location of user

Page 10: Stefan Marti Speech Interface Group MIT Media Lab.

Related DocumentsRelated Documents

Main feature that makes it better than other solutions: it can wait!

Scenario: What happens when an email message arrives?Let’s assume the following “channel sequence”:

• wait 10 minutes, send to Pager• wait 13 minutes, send to Voicepager• wait 13 minutes, send to Phone• wait 35 minutes, send to Fax

Page 11: Stefan Marti Speech Interface Group MIT Media Lab.

Monday 6:50am 7:00am 7:10am 7:20am 7:30am 7:40am 7:50am 8:00am

Ok to usePager: anytime

Voicepager: not 0-7

Phone: not M-F 22-8

Fax: not 2-7:30

6:57am arrival of message

send to Pager10 minutes

send to Voicepager13 minutes

Ok to use

send to Phone

13 minutes

User reads message!

35 minutes

Ok to use

NOT ok to use!

done!

done!

skip it!

send to Faxcancel!

Page 12: Stefan Marti Speech Interface Group MIT Media Lab.

How Does It Work?How Does It Work?

How does AM choose specific “channel sequences”?

AM solves the routing and forwarding problem using context information.

Page 13: Stefan Marti Speech Interface Group MIT Media Lab.

Context SensitivityContext Sensitivity

1. User location Where is the user?

2. Message priority How important is message?

3. Message history Is this message part of thread?

4. Message status Was the message read?

5. Channel status Which channels are active?

Page 14: Stefan Marti Speech Interface Group MIT Media Lab.

User LocationUser Location

Where is the user?

Most devices are not location aware

Obvious: AM keeps track of user with caller ID information and UNIX “finger” command

AM also infers from user behavior: From which channel did she communicate most recently?

Soon: cell trilateration, other tracking options like GPS. (Will make AM just more accurate.)

1. User location

2. Message priority

3. Message history

4. Message status

5. Channel status

Page 15: Stefan Marti Speech Interface Group MIT Media Lab.

Message PriorityMessage Priority

How important is a message?

Complex problem – almost philosophical. What is really important in my life?

Anyways: Certain people are important, certain facts too:

Static rules

AM uses other information sources, like user’s calendar, address book, and communication history:

Dynamic rules, created by CLUES

1. User location

2. Message priority

3. Message history

4. Message status

5. Channel status

Page 16: Stefan Marti Speech Interface Group MIT Media Lab.

Message HistoryMessage History

Is the message part of a thread?

Instead of sending a message to a sequence of channels, pick the channel it originates from.

This assumes that the message is part of a thread, going back an forth, perhaps a kind of semi-synchronous communication.

1. User location

2. Message priority

3. Message history

4. Message status

5. Channel status

Page 17: Stefan Marti Speech Interface Group MIT Media Lab.

Message History, cont.Message History, cont.

Usually: Channel sequence

1. User location

2. Message priority

3. Message history

4. Message status

5. Channel status

Option: Association of channel with thread

Page 18: Stefan Marti Speech Interface Group MIT Media Lab.

Message statusMessage status

Is message read?

When message is read, job is done!

Usually not enough feedback from devices

BUT: User’s behavior could give a hint!

Message read level is not binary (yes/no), but a continuum, a “probability level” between 0% and 100% (85% is enough usually.)

1. User location

2. Message priority

3. Message history

4. Message status

5. Channel status

Page 19: Stefan Marti Speech Interface Group MIT Media Lab.

Channel StatusChannel Status

Which channel is active?

AM continuously checks all peripheral systems if they are working

It skips channels and devices that seem to be not working

Again, it uses back channel information from user

1. User location

2. Message priority

3. Message history

4. Message status

5. Channel status

Page 20: Stefan Marti Speech Interface Group MIT Media Lab.

Summary of AM FeaturesSummary of AM Features

Context information is used to solve the routing and forwarding problem

Behavioral heuristics are applied to enhance unclear context information

AM has a temporal dimension: It can wait for user reactions and other events to happen

Page 21: Stefan Marti Speech Interface Group MIT Media Lab.

Stefan Marti and Chris Schmandt{stefanm|geek}@media.mit.edu

Speech Interface Grouphttp://www.media.mit.edu/speech