Subject: RESCUE MEETING
Date: 01/30/04
Loc: URP
Author: AM
SM: Demo status.
Data set status.
Dissemination
Camas1 – Dmitri
Surveillance – Jehan/Mahesh
Filtering – Qi/Yiming
Alternate 911 – Kemal.
Reliability - Fabio
DK: *AM to improve extraction in Camas.
AM: Facilities data available. Not extraction so much as normalization.
SM: [Summary of 2-day-earlier meeting with UCI ICS Helpdesk, NACS, Facilities, EMC]
NV: Improve problem tree.
Use data as real input.
SM: Classification on NACS, ICS Helpdesk.
On email reports.
Learning from pre-classified email set.
Not speech, no word-spotting.
Adaptive models.
DK: Borrow Smyth algorithms.
SM: Somewhat different problem from ours.
Smyth is extending LSI (Latent Semantic Indexing).
We: interactive learning from misclassification.
Could use Smyth work as a starting point.
DK: Names the classes with concatenation of keywords.
Could interactively rename.
SM: *Could get their code.
- Classification.
- Interesting: set of attributes, auto extract values from the text itself.
Text-based
- Go from voice/speech
*DK, Monday, scheme for doing this end-to-end system.
Raw data to event representation.
Initial thoughts, design, flow diagram, instantiation, example model.
?: This is research, may or may not work.
SM: March 19 site visit.
Carter found transcripts of 9/11.
DK: *Someone to beautify the Camas web pages.
SM: *Separate Camas1 demo plan; assign tasks.
*Submit plan to SM, CB, NV, PS for feedback.
Jehan: [Surveillance Status]
Hardware order status, in progress.
Crypto techniques.
NV: Model an airport.
Rooms, video cams.
All: [Filtering Status]
[Event Extraction]
End to end.
Jehan: Speech architecture already at UCSD.
SM: Time is short, but use it in the future.
Integrate with theirs.
Like the AT&T HMIHY system.
They built traffic advisory via speech interface.
Interactive navigation through help space.
Simple, stupid.
*We want to incorporate, extend speech technology itself.
UCSC focus is noisy environment.
CB: Don’t want to step on each other’s toes.
Divvy up tasks, else turf wars.
SM: Shino commonality. Can we use?
No privacy.
Not asking us for products, results.
Linda Bogue – UCI Emergency Management.
70 emergency workers at UCI.
NV: Simulated drill.
SM: They’re open to working with us on drill, UCI-wide.
*CB is point of contact.
CB: This is an argument for instrumenting the campus during drills.
Help reconstruct what happened.
To help with future drills.
By March:
(1) Difficult for her; we develop.
(2) Letter of support.
(3) Phyiscal presence – ok with Linda.
Eval what happened.
SM: Leery of their plan, imagination.
May be copied from elsewhere.
NV: Little dependence on technology.
CB: (1) Tech for studying what they do.
(2) Give them the tech later on.
SM: Not integrated yet, we can do this.
Real, integrated university drill.
Field deployment.
SM: [Dissemination Research]
Xingbo.
Peer-to-peer dissemination papers.
Mayur.
Proposal.
(1) Develop system fro contextualized dissemination.
- To field workers.
- (main focus) To the public.
First and second-hand accounts.
Watching news, cellphones on.
Estimates of channels, tech, logins.
Design contextualized dissem solutions for this scenario.
CG: NTSC (?) report on effective dissem solutions.
Need a standard.
Electronic – on website.
And people who don’t have access.
Sirens!
Selectivity – else warnings are ignored.
All: Champagne storm – siren didn’t work.
SM: Security authentication, trust issue here.
NV: ABC networking – Always Best Connected.
Cellular, wireless LAN, Bluetooth…
Seamlessly use the best that’s accessible.
Chen: Highway board.
?: Individualized maps disseminated, eg.
SM: Dynamic creation of peer-to-peer network.
NV,CB: Barriers.
Language
Disability
Technology availability.
SM,CB: Cellphone companies – demographic information.
Privacy preserving.
Xinbo: [FUDAN][PEER SDI]
Selective Dissemination Information.
SIFT – Stanford U system, example.
All: [Xinbgo continued]
[Trying to understand the model]
[Mapping dissemination model to handle disaster]
[Discussion…]
What goes into a profile.
What is disseminated.
SM: [Whiteboard]
[
Dissemination
1. Within emergency response network
a. Peer to peer
b. Commands, downstream echelons
2. Public
a. Group
b. Individual
Naming -> verification
3. Resource discovery through network
]
GS: United.
Radio.
CB: Old phone tree.
NV: *Dissemination group?
[Note taker leaves 4:15 – AM]
All: Presentation from each group, 2 weeks from now.
CB: Contact Linda Bogue. (AM give ctc info – or Google ;-)
DK: Monday, scheme for an end-to-end system. 1 page writeup.