Integrating and publishing public safety data using semantic technologies

Integrating and Publishing Public Safety Data Using Semantic Technologies

Alvaro Gravesalvaro@graves.cl

Tetherless World

Constellation

Department of Cognitive Science

Outline

Motivations

Implementation

Challenges and next steps

Conclusions

<Motivation>

<Research>

Research Motivation

Lots of “human computational power” An example:

– 9 billion human-hours of solitaire were played in 2003 (in red)

– Building Panama canal took 20 million human-hours (in blue)

Interesting, but...

… How can we use this cognitive and computational processing power...

… to solve difficult problems....

… (Ideally) without much effort from the users?

Social Machines

• Social Machines are mechanisms where:

– Humans do the creative work

– Machines do the administrative work

What are we good at?

Humans MachinesPattern Discovery Good BadCreative Thinking Good BadInformation Management Bad GoodData Communication Bad Good

Example 1: reCAPTCHA

Example 2: GalaxyZoo

Example 3: Threadless

How to study Social Machines?

Limitations Access to information (logs, database) Privacy concerns Lack of flexibility

Solution: Create framework for Social Machines Incentives are important

</Research>

<Practical>

Troy, NY

Segmented city– Downtown

– RPI

– North/South

High risk zones

Unlike big cities– Hard to have

centralized data

– Resources are scarce

How can... Citizens to be aware of their environment?

– How risky is to park in this street between 1AM and 3AM?

Policy-makers make right decisions?– Is a specific crime increasing over time?

Law enforcers be more transparent in their work?

– How to show our work to the community?

What can we do?

Idea: Let's build....

… a platform for integrating public safety information

… a framework for studying Social Machines– Understand behavior of users in front of

data– Run experiments (ex: provenance => trust)

</Practical></Motivation>

<Implementation>

Architecture

<SemanticWeb>

Semantic Technologies

Based on the Web– Don't need to create special platform

Domain agnostic– Can express different domains

Dynamically coupled– Easy to mix different data sources

(Also: Interoperable, distributed, extensible, etc.)

Semantic Web: Example

Most basic “language”: RDF (Resource Description Framework)

Set of assertions (triples)

My name is Alvaro Graveshttp://graves.cl#me foaf:name “Alvaro

Graves”

Add more triples

</SemanticWeb>

<DataIntegration>

Different sources...

RPI Public Safety PDF files available on the Web Different formats depending on which

year Not easy to extract data

Troy Police Department Excel files upon request Less information (date/time, event type,

geographical references) Only certain crime types

...Similar Problems

Unclear geographical references (“Dining Hall”, “15th and Peoples ave.”)

Not only crimes but other events (fires, medical emergencies, etc.)

Our Taxonomy

</DataIntegration>

<Curation+Persistance>

Existing data is not enough

Lack of geolocation Use Google Maps as a Web Service

for obtaining latitude and longitude Typos

Manual correction Ambiguity

“Event occurred Off campus” (!) Approximate to the best we could

An example

Persistence

Use of semantic technologies for persistence

Pros: Easy to setup Works with MySQL+PHP

Cons: Not very scalable Queries cannot be too complex

</Curation+Persistance>

<Visualization+Publishing>

Data in multiple ways Citizens

At home/office On the street

Decision makers Statistics Potential correlations?

Machines Create notifications Display in other platform Reuse data

PublicSafetyMap.org

PublicSafetyMap.org

m.PublicSafetyMap.org

Only data around your location

Still in development!

PublicSafetyMap.org/feeds.php

Latest data Current search

RDF RSS KML

</Visualization+Publishing></Implementation>

<Challenges>

More than technical problems

Hard to convince decision makers

Trust issues

“Social entry barriers”

Next steps

Connect with well-known social networks (Facebook, Twitter, etc.) Geolocate anything (Bicycling route, pizza delivery area)

– Geoinferencing Add more dataset (Weather, real estate, Fire dept.)Annotation on events (Establishing trust measures, explanations, etc.)

Next steps

</Challenges>

<Conclusions>

Research: Studying Social Machines

Don't compete with Google

Opening data so others can use it for their own applications

But most important: A Framework where we can study Social Machines

Being able to run experiments

From Practical:Public Safety Information

User can visualize their neighborhood.

Policy-makers can manage data in a useful way.

Law enforcers can show their efforts.

</Conclusions><mailto:alvaro@graves.cl/>