October 24, 2007

GEOINT 2007 Symposium - San Antonio, TX

SSS Research Announces Collaboration with NGA

(National Geospatial-Intelligence Agency)

Visual Intelligence (VI)^TM Software Enables Location Intelligence For Rapid Response and Coordination

San Antonio, TX – October 24, 2007 – SSS Research today announced that it is working with the National Geospatial-Intelligence Agency (NGA) to provide key software as part of its NGA Earth eGeoInt project. Under this relationship, SSS Research provides its GeoBoost Visual Intelligence^TM software solution to enable situational awareness and location intelligence for multiple constituencies.

“We are very pleased to have been selected to provide the visualization software for this key initiative,” said Dr. Stephen Eick, SSS Research’s Chairman and CEO. “NGA plays such an important role in supporting the Intelligence and Defense communities, we are happy to collaborate on such a critical analysis tool.”

One of NGA’s 12 key focus areas is to look outward to be the most collaborative partner with the IC and warfighter. To support that focus, NGA has identified 3 key principles: Partnerships, Technology, and Standards. The innovative technology developed and deployed by SSS combines with a successful track record of previous projects with NGA to deliver a powerful analysis and visualization tool for the Intelligence Communities.

Before the emergence of Web 2.0 technologies, organizations were limited in their ability to leverage Geospatial Information Systems (GIS) and other location-based data. SSS Research has leveraged Web 2.0 (AJAX) and open data standards to enable geospatial visualization, collaboration, and analysis to users through a software solution that requires only a web browser. The result is better insight, faster across the entire enterprise in an easy-to-use tool that connects to multiple data sources simultaneously using open standards. A recent study indicated that 80% of data maintained by organizations has a location component; and the demand to access and analyze these data continues to grow. With the flexibility offered by SSS Research’s suite of VI^TM applications, organizations now have the ability to turn this data into actionable analysis quickly and easily.

About SSS Research

SSS Research has served the Federal IT marketplace since 2003 across many departments and agencies of the United States Department of Defense.  Core competencies include thin client visualization and analytics R&D services. SSS has developed a unique Visual Intelligence(VI)^TM product set – GeoBoost and GeoTrack - which bring together the power of geospatial visualization technology and business intelligence into a thin-client SOA-compliant platform.  Simply put, we help our customers visualize their data in ways never before possible, helping them make better decisions.  The Web 2.0 platform allows this solution to integrate with any data source and be deployed and interacted with in real time on any thin-client via browser, including hand-held and mobile devices. 

Core commercially-available products include:

• GeoBoost - a Business Intelligence application - combines spatial or geospatial data with business intelligence (BI) data to enable location and time-specific analysis in a unique visualization portal. 
• GeoTrack - an asset visualization solution - interacts with RFID (Radio Frequency Idenfication), GPS,  or other data and provides a visualization and business rules solution for managing critical assets or personnel, location intelligence and situational awareness.

July 16, 2007 - SSS Research has moved to a  new office located at 801 Warrenville Road, Suite 50, Lisle, IL.  630.596.5420 main phone 630.596.5443 efax

Naperville, IL  ...............Announcing! Dr Stephen Eick, President and Chief Scientist of SSS Research has been selected to perform as a Primary Speaker at the Intergraph 2007, our annual flagship users' conference.  Intergraph 2007 is honored to have you as part of our conference agenda.

Session Track: Geospatial Data Collection, Sharing, and Analysis

Title: Interoperability/Sharing: Creating Online Web Mapping Applications with GeoBoost

Date & Time: May 24 2007 10:00AM - May 24 2007 10:45AM
Room:  Lincoln E

Session Abstract: GeoBoost is a Web 2.0 client-collaborative Web mapping and visualization application that brings a fresh and powerful front-end to users’ existing GeoMedia WebMap and TerraShare applications. The solution provides an AJAX-based streamlined interface, enhanced graphical user interface, thin client visualizations, real-time geospatial collaboration, and increased flexibility for data access and integration. This session will also detail the implementation of the GeoBoost technology from SSS Research and the mechanism for integration into a TerraShare and GeoMedia WebMap architecture. This session will show how GeoBoost, serving as the client application, integrates with GeoMedia WebMap as the geospatial processing engine and with TerraShare as the image archive management engine. Elements of integration via Open Geospatial Consortium (OGC) open Web mapping services will also be covered.

Biography: Dr. Stephen Eick serves as deputy director for the National Center for Data Mining at the University of Illinois in Chicago. He is the President and Chief Scientist for SSS Research, a company that he founded in 2003, whose focus is thin client geospatial technology.

Naperville, IL.   SSS Research announced today that Kurt Rivard will be presenting and providing a posterboard presentation on Wed May 16, 2007 during the 6-8pm reception located at the Holiday Inn Select BOSTON-WOBURN, Woburn, MA. He is providing a slide deck show illustrating :

A Web 2.0 Framework for Real-time Location Intelligence

Next Generation Geospatial Information Visualization and Collaboration

•Web 2.0

•Our Web 2.0 Approach

•Crisis Management Needs

•How our Platform addresses those Needs

Show Information:

The Seventh Annual IEEE Conference on Technologies for Homeland Security will continues the tradition of providing a unique venue for the cross-fertilization between operational expertise; governmental and other organizational requirements and novel technologies and technical solutions. The Conference will focus on technologies ranging from devices and computer software to integrated systems with the goal of deploying innovative solutions enhancing the security, efficiency and reliability of transportation, communications, power and other critical infrastructure.

Through plenary and breakout sessions, exhibits and special events, conference participants will:

• Address the most pressing technical challenges;
• Identify the most critical knowledge gaps;
• Explore the application of advanced sensors in transportation systems;
• Understand opportunities for small businesses to work with government and large system integrators.
• Network with others involved in critical infrastructure issues

Background

The seventh of the IEEE Conferences on Technologies for Homeland Security is sponsored by the Boston Section of the Institute of Electrical and Electronic Engineers and the IEEE Critical Infrastructure Dependability Initiative and organized by the Boston Homeland Security Conferences Steering Committee, Len Long, MSEE, CPP, qp, Chair.

The conference, business panels and technical expo provides an ideal opportunity for developers and implementers of technologies, consultants, entrepreneurs and business funding sources interested in the security and critical infrastructure dependability marketplace to meet with governmental leaders and owners and operators of major private sector infrastructure.

Kurt will be providing a posterboard presentation on Wed May 16 during the 6-8pm reception.

2007 IEEE Conference on Technologies for Homeland Security:

Enhancing Critical Infrastructure Dependability

Wednesday May 16 and Thursday May 17

Holiday Inn Select BOSTON-WOBURN, Woburn, MA

Event link

October 21st - 24th

Gaylord Opryland Resort and Convention Center

San Antonio, Texas

Event link

Booth # 13

Gaylord Opryland Resort and Convention Center

May 21st-24th

Nashville, TN

Event Link

April 16th - 18th

Parc 55 Hotel San Francisco, CA

SSS Research - Booth # 14

See  - Keynote Speaker:  Andrew Eick, CTO of SSS Research, Inc.

Wednesday - April 18th at 2-3pm in the DaVinci Room

Topic of Session:   Andrew will be discussing AJAX SOA and other Web programing techniques such as GeoRSS to provide a rich dynamic user experience in a Web browser.

 event link

A Web 2.0 Framework for Real-time Location Intelligence

Kurt Rivard, SSS Research, Inc.

Printable View

Abstract – With increasing deployments of Global Positioning System (GPS) devices, Radio Frequency Identification (RFID) tags, and other location-aware devices, it is now possible to capture time-varying object information. In addition, existing systems such as Enterprise Resource Planning (ERP) systems emit ancillary object metadata (e.g., inventory levels in a warehouse). The challenge is how to leverage these information assets for demanding field operations scenarios such as Crisis Management, given the size and real-time nature of the data. To address the challenge, we have built a Web 2.0 framework for real-time spatial intelligence and collaboration. Our framework includes an extensible architecture for ingesting and combining spatial data across multiple formats; a fusion server for merging spatial and bespoke business data; support for spatial transformations tied to configurable business rules; and a publishing engine that pushes the combined information out for consumption in a visual, collaborative presentation layer running in standard Web browsers and on mobile devices. The result is an easily deployable system with broad reach to the field through a visual, interactive interface presenting timely, meaningful information.

Introduction

Crisis Management requires fast, on-the-spot decision-making by emergency responders (ERs) addressing reported incidents. Establishing a usable Common Operating Picture, or Shared Situational Awareness, for all the involved teams and individuals, possibly cross-jurisdiction, in both crisis and routine situations is essential to successful outcomes.

Fundamental needs of ERs surrounding Shared Situational Awareness include getting critical information related to the crisis at hand in real-time in an easily digestible form for fast understanding; enabling quick, intuitive interrogation for further detail; being alerted to unusual or highly important events related to the crisis; easily and securely communicating location-based observations and information with some or all involved parties; and easily accessing the system being used from standard devices without needing to manage client software.

Systems in the past have failed to meet the challenges posed by these needs because they were too complex for the targeted user (ERs), too limited in capability, relied on the deployment of specialized software, were too expensive to license and/or maintain, or some combination of these reasons.

To address the challenges of providing Shared Situational Awareness for Crisis Management, we have developed a Web 2.0 framework for real-time spatial intelligence and collaboration. Our framework is unique for several reasons. First, the architecture is modular and built around open Web and geospatial standards. This open architecture simplifies integration, makes it easy to extend, and enables “bottleneck tuning” to support varying deployment scenarios (e.g., bandwidth constrained environments). Second, the publishing engine has the capability to generate output customized for consumption on various devices. These devices include thin client Scalable Vector Graphics (SVG) [1] and Asynchronous JavaScript and XML (AJAX) [2] enabled Web browsers and mobile devices. AJAX and SVG enable a rich client experience without the need to manage client software. Third, our framework includes a novel presentation layer with custom visual displays for geo-fencing, geo-tracking, bread crumbing, hot spot analysis, collaboration, and location prediction – all through thin client and mobile device interfaces.

In this paper we describe the system in the context of Crisis Management. Specifically, we provide an outline of ER needs surrounding Shared Situational Awareness followed by a description of our framework capabilities that address those needs. We conclude with a summary of benefits.

Emergency Responder Needs

Crisis Management involves one or more teams of people and various support resources (e.g., emergency vehicles, command and control unit, radios, and medical equipment). People may be in the field (e.g., policemen, firemen, emergency medical technicians, HAZMAT responders, on-site incident commander) or off-site behind the scenes (e.g., dispatch, emergency operations center). While we refer to the people in the field as emergency responders (or sometimes “first responders”), the needs of a system supporting Shared Situational Awareness extends to the people working behind the scenes, particularly at the operations center.

ER needs for Shared Situational Awareness support can be categorized into usability needs and system needs. First we identify the usability needs:

(1)

While ERs need to be presented with all the facts, it is essential the information is presented in a form that they can quickly internalize within the context of the present situation. This suggests the supporting system provide an intuitive, visual approach that enables ERs to glean immediate insight about the situation simply by glancing at the information display.

(2)

ERs need to know the “what”, the “where” and the “when” about relevant events and objects. Understanding how a group of events are related across both location and time contributes to improved Situational Awareness. This suggests the supporting system present the information (the “what”) in both geospatial (the “where”) and temporal (the “when”) forms from which associations can be derived.

(3)

ERs need to know where and when critical events related to the crisis have occurred, and where new related events occur as they happen. This suggests the supporting system provide well-defined data ingest protocols and efficient, robust data management.

Further, knowledge of current positions of emergency vehicles and personnel that have responded or could respond to an incident is essential in the coordination of the overall effort. A recent history of where a responder has been is useful for understanding things like surveillance coverage. These needs suggest the supporting system provide the ability to represent multiple types of data (e.g., events, objects, paths).

(4)

ERs need quick access to more detail than what might be initially presented. Detail might be specific to the event itself (e.g., type of chemical spill) or contextual information surrounding the event (e.g., see aerial photography of the affected site at a granular level). This suggests the supporting system provide an interactive interface that allows the user to interrogate and interact with the displayed data.

(5)

ERs will have on-the-spot observations that will be pertinent to the rest of the team who are both in the field and in the operations center. Most of the time location will be a key element of those observations. The ER will need an easy way to identify/describe the location – typically an area – and make others aware of it, and perhaps include commentary in some form capturing key observations. This suggests the supporting system provide built-in location-based collaboration.

(6)

ERs want to be told when something of high interest occurs – particularly something that deviates from the norm – that might be related to the crisis, e.g., a looting crime occurring in or around an evacuated area, or an emergency vehicle leaving or entering a designated quarantined zone. This suggests the supporting system provide a flexible, rules-based approach for triggering predefined actions associated with events and object positions relative to designated zones.

System needs are as follows:

(7)

The system that supports the above usability needs must be readily and easily accessible by ERs, operations center personnel, and other parties involved in the management of the crisis. The system must support common client devices with standard environments without imposing specialized software requirements. Updates and/or upgrades to the system must not require changes to the client environment. This suggests a Web-based, thin client system that requires no software downloads or plug-ins.

(8)

The system must easily integrate into existing environments/applications. This suggests an open system based on accepted, industry standards.

Web 2.0 Approach to Crisis Management

Our approach was to build a framework on Web 2.0 from the ground-up to support real-time spatial intelligence and collaboration and a thin client deployment. Core capabilities as they relate to the Crisis Management needs identified in the previous section are described below.

(1) At-a-Glance Understanding

Figure 1 shows a simple Crisis Management application built on our framework. The map provides location intelligence of a number of 911 calls, including the crisis incident (HAZMAT). The translucent orange area on the map represents the affected area to be evacuated, as indicated by the tear-away tooltip shown.

The Timelines at the bottom provide a temporal view of the items shown in the map. The top Timeline provides a summary view of all notable events that have occurred so far throughout the day. The bottom Timeline provides a detailed hourly view of a subset of the day represented by the blue lens in the top Timeline, roughly 10:30am to 2:30pm.

Figure 1: This sample Crisis Management Application provides at-a-glance insight to support Situational Awareness.

The map and Timeline provide an intuitive understanding of where and when events occurred. The icons themselves provide contextual information about the type of event (i.e., HAZMAT versus petty theft versus assault). Note that a given item is represented with the same icon in the map and Timelines. This helps to see not only what kind of events occurred, but which events occurred when.

(2) Relating Location and Time

The map and Timeline are linked. Pointing your mouse at an item in the map highlights it in the Timeline, as shown in Figure 2, and vice versa, and shows detail in a pop-up tooltip.

The association between location and time can be seen across multiple items by sweep-selecting a group, as shown in Figure 3. We can see quite clearly that three of the four events (which had close proximity to each other) we swept over in the map occurred very close in time.

Figure 2: Get detail and see association of an item's location and time by pointing your mouse at it.

Figure 3: Linked Map and Timelines enable ERs to easily associate event location and time.

We also employ automatic “aging” of events whereby the points/icons gradually fade with time (and possibly other factors), The points eventually fade completely from the display. This is used as a way to manage data cache size with real-time data feeds as well as provide a visually effective way to relate location and time.

(3) Real-Time Events and ER Positions

In the sample application shown in Figure 4, items are automatically being ingested into the system. The user does not need to manually refresh the data.

Behind the scenes, event data is being generated through 911 calls and stored in other existing databases. Our client supports the notion of a Data Connector, which in this sample application is configured to access the framework’s Tracking Server through its Web Service interface to get at the 911 event data. The Data Connector polls the Web Service at regular intervals for new events (the data could also be pushed to the Data Connector). It then intelligently adds the new events to the data cache, possibly rolling off old events that have expired based on application settings.

In Figure 4 there are two types of items being displayed: events (911 calls) and objects (current ER positions). These are represented as separate “data overlays” on top of the map and can be toggled on or off independently via the navigation panel at the left.

Behind the scenes, ER vehicle locations originate from a Global Positioning System (GPS) source. The GPS data is ingested through the framework’s Tracking server. Contextual information about the items, such as the squad car name and the icon being displayed, is ingested from a different system, fused with the location data, converted to GeoRSS, and sent to the client.

Only the latest reading for each ER vehicle is displayed. However, the recent positions can be shown in the form of a “bread crumb” trail attached to the current position. This is shown in Figure 4.

Figure 4: Objects (Emergency Responder positions) are displayed along with events. Their positions can change in real-time. "Bread crumbing" (pathing) is used to show where the objects have recently been.

Finally, note that only events are shown in the Timeline. The ER vehicle readings are not. This is because we are interested only in the current positions and where they have recently been. The timestamp for these readings are uninteresting since they are all taken at predefined intervals.

(4) Detail on Demand

Our Web 2.0 framework provides a rich user experience. There are a number of built-in interactions the user can leverage to see more detail or additional context such as

• Pan and/or zoom the map(s) (multiple maps can be leveraged with synchronized panning and zooming).
• Pan and/or zoom the Timelines (Timelines support synchronized panning, so in our example, as you pan the top Timeline, the bottom Timeline automatically adjusts to show events positioned within the blue lens).
• Zoom to a selected set of items.
• Allow multiple map layers to be overlaid on the display.
• Add more (raster) detail to the map through the selection of additional map layers (this will be dependent on the WMS source being used).
• Adjust the transparency of the top map to see the bottom map (e.g., blend street map with close up aerial photography to get a better sense of the structures and land use of the area surrounding the incident).
• Show/hide other available data overlays like public transportation routes and stops.
• Display items as points rather than icons.
• Change the color of the events (e.g., to reflect priority/importance).
• Click on “More Detail…” in an event’s context menu to pull up a Web page providing up-to-date detail on the event and background information about the hazard.

In Figure 5, we have zoomed one level down to the area where the HAZMAT incident occurred. We then brought up the linked mini-map, which provides a magnified aerial photography view of the area represented under the blue lens in the street map. The aerial view provides useful context for ERs.

Figure 5: ERs can interact with the Map, Timeline and displayed data to get at more detail on demand. Synchronized panning and zooming are standard interactions. In this illustration a mini-map is synchronized with the street map. It is used to provide a magnified aerial view of the area positioned underneath the blue lens of the street map.

Behind the scenes¾Our framework leverages Web 2.0 technology, which is largely asynchronous JavaScript and XML, also referred to as AJAX. With AJAX, information is asynchronously downloaded to the browser, cached by the JavaScript, and displayed by the client when the user requests it. The net effect is an instantaneous response, without a disruptive page refresh. This makes possible a direct manipulation user interface, strongly preferred by users [3], but not possible with traditional (Web 1.0) applications due to the latency between the request and response and the disruptive nature of the page refresh [4].

(5) Communicate Location-based Information

The orange polygon in our example represents the initial evacuation area that was created by an ER or someone in the operations center shortly after assessing the crisis incident earlier in the day. Figure 6 shows the area being reconfigured based on a new assessment. As shown, the polygon is drawn with the mouse or stylus (depending on the user’s device). Further, the user can attach commentary in the form of notes or audio recordings.

Figure 6: ERs can mark the map directly and attach observations.

Markings and attachments are automatically propagated to other ERs by our framework’s Collaboration Server, as illustrated in Figure 7. User-drawn zones are rendered as a separate data overlay. ERs have the option to show/hide these zones.

Figure 7: Markings automatically propagate to other ERs' displays.

(6) Alert Me to Important or Unusual Matters

Figure 8 shows an alert triggered by an ER vehicle entering the quarantined zone. The alert was triggered based on a business rule previously set up to kick off an alert whenever an ER vehicle enters or exits a quarantined zone. In other words, even zones created on the fly by ERs will automatically be subjected to pre-defined business rules.

Figure 8: ERs are alerted to important occurrences based on configurable business rules.

Our Tracking Server provides Business Rules and Alerting modules to define rules and actions, dynamically perform calculations as location data is received, and trigger corresponding alerts.

(7) Broad Reach (Thin Client)

Our Web 2.0 approach enables rich client interaction while maintaining broad reach through a thin client interface that imposes no specialized software requirements on the client device.

(8) Easy Integration

Our framework supports a Representational State Transfer (REST) Service Oriented Architecture (SOA) allowing easy integration into existing environments and applications. In addition, it provides flexible, standards-based imagery and feature data ingest subsystems.

Imagery¾Our frameowrk supports the OGC standards WMS and WFS. It also is able to read imagery formats like RPF (CADRG and CIB), MrSID and JPEG2000. Further, it provides a WMS interface to all these sources, including other WMS’s, enabling a single view to consist of layers from multiple image sources. This flexible imagery ingest enables imagery sources to be hooked up with little effort.

Data¾Our framework supports standard protocols for feature data ingest including GeoRSS [5] and Geospatial Markup Language (GML) [6]. GeoRSS extends the popular Really Simple Syndication (RSS) protocol [7] with location information. Standard GeoRSS supports points, lines, circles, and polygons. Our framework also supports thincGeoRSS, an extension to GeoRSS that supports additional shapes like ellipses and the specification of common attributes like fill color, line thickness and transparency. In addition, our framework can ingest JavaScript Object Notation (JSON) [8] as well as JavaScript itself. Supporting standard protocols as such simplifies the process of connecting up data.

Client¾Our framework imposes no restrictions on the client except requiring Scalable Vector Graphics (SVG) full version for client-side rendering of feature data. SVG is supported natively in Firefox, and as a plug-in (from Adobe) for Internet Explorer. Although our framework currently supports only SVG for the rendering, it leverages a rendering abstraction layer, making it easy to support other rendering engines such as Microsoft VML or XAML.

Conclusions

A usable Shared Situational Awareness is crucial for successful Crisis Management. The high-pressure, stressful conditions under which ERs must operate makes it imperative that the tools they use are easily accessible, reliable, and intuitive. Our thin client approach provides reach (accessibility) and does not affect the client environment, increasing reliability. GeoBoost’s novel, highly visual displays present the information in a format digestible by the ER at-a-glance. ERs are informed of critical events through custom alerts triggered by business rules tailored for Crisis Management.

Getting at the right information as events occur and sharing key observations are also crucial to Shared Situational Awareness. Our framework was designed to deal with the complexities of efficiently handling real-time data. It was built on Web 2.0 technology from the ground up, providing a rich environment for on-the-spot interrogation of the information and its surrounding context. It also enables ERs to share location-based observations simply by marking/annotating the map, with backend support to propagate those observations to the involved ER teams.

Finally, our framework’s flexible architecture and standards support make it easy to integrate into existing environments.

Simply put, the capabilities of our framework are well-aligned with Crisis Management needs. One challenge at hand is supporting handheld devices, like SmartPhones and PocketPCs, where SVG and other standards are still ill-defined. Our current approach is to provide client-side code specific to the device, and then revisit a true thin client once standards on those devices become more settled.

Our Web 2.0 approach coupled with visualization and strong server-side functionality offers a number of key advantages surrounding reach, ease of use, richness, and integration that lend itself to Crisis Management and many other problem domains.

References

[1] Scalable Vector Graphics (SVG) 1.1 Specification, W3C Recommendation, 14 January 2003, www.w3.org.

[2] Building Rich Web Applications with Ajax. IEEE Computer 2005; 38(10): 14-17.

[3] Designing the User Interface, Addison Wesley, Third Edition, 1998.

[4] Stuart Card, Jock Mackinlay, Ben Shneiderman, Readings in Information Visualization: Using Vision to Think, Morgan Kaufmann, 1999.

[5] Encoded Objects for RSS feeds, April, 2006, www.georss.org.

[6] OpenGIS® Geography Markup Language (GML) Encoding Specification, OGC 02-023r4, Version 3.00, 18 December 2002.

[7] RSS 2.0 Specification, blogs.law.harvard.edu/tech/rss.

[8] Introducing JSON, RFC 4627, www.JSON.org.

Source

IEEE Symposium on Visual Analytics Science and Technology, 2006

Authors:

Stephen G. Eick
SSS Research, Inc.

Justin Mauger
SAIC Advanced Systems & Concepts

Alan Ratner
National Security Agency

Abstract

We have built a visualization system and analysis portal for evaluating the performance of computational linguistics
algorithms. Our system focuses on algorithms that classify and cluster documents by assigning weights to words and scoring each document against high dimensional reference concept vectors.  The visualization and algorithm analysis techniques include Confusion Matrices, ROC Curves, Document Visualizations showing word importance, and Interactive Reports. One of the unique aspects of our system is that the visualizations are thinclient web-based components built using SVG visualization components.

CR Categories: H.5.2 [Information Interfaces and Presentation]:
User Interfaces—Screen Design; H.3.4 [Information Storage and
Retrieval]: Systems and Software—Performance Evaluation;
[H.3.1 [Information Storage and Retrieval]: Content Analysis and
Indexing—Linguistic Processing; H.3.3 [Information Storage and
Retrieval]: Information Search and Retrieval—Clustering

Keywords: AJAX, thin-client, SVG, ROC curves, confusion
matrices, document categorization

1 INTRODUCTION
One of the current challenges in computer science is developing new methods to represent, structure, analyze, and automatically process unstructured multilingual text to obtain semantic understanding. To aid with this task, we have built an analysis system and portal that helps researchers quantify the performance of their algorithms. Our target users are developing hardwareaccelerated algorithms that are described in other papers, see for example [1], [2], [3], and [4].

The goals for our analysis system, called AFEWeb, are to provide unambiguous performance metrics for algorithm
developers, develop visualizations that show key algorithm abstractions, and create new visualizations that show why
documents are identified as belonging to specific concepts of interest. Our system compares the performance of various algorithms, helps our algorithm designers set thresholds, provides simple and intuitive access to misclassified documents, and displays key algorithmic concepts to users and developers in an understandable way. It cross-references all of the words seen, it is fully interactive, and displays in a standard web browser.

AFEWeb includes many standard document algorithm analysis tools such as Confusion Matrices, ROC curves, Precision and Recall plots, and Interactive Reports. It also includes new visualizations showing document clusters, and visualizations of the words in the documents themselves. Although, some of the analysis tools are known, what is new is to have them packaged together as an analytical system within a web-based document analysis portal. Furthermore, some of our visualizations are novel and particularly well-suited to document analysis. Our interactive portal is surprisingly useful. Using our portal we have identified several algorithm bugs and algorithm performance characteristics that have influenced the developers and our corpus collection techniques.

Our portal is written using AJAX2 , SVG3, and other Web 2.0 programming techniques. Using these techniques, it is possible to develop rich interactive user visualizations that are totally browser-based. Our portal combines the utility of rich desktop analysis systems with the flexibility and linking of web-based systems. It works remarkably well and demonstrates that is possible to build powerful visualization tools in a thin-client browser platform.

In the remainder of this paper we describe our system in more detail. First, however, we review key concepts for processing unstructured information using the constraints imposed by our hardware testbed as a concrete example. In subsequent sections we will describe AFEWeb’s analysis tools and interactive portal.

Read full paper

A case study using DECIDE for intelligence analysis.

Plots

Hypothesis: Political corruption is present in the Alderwood government.

Political corruption seems to be present in several forms: election tampering, unethical behavior, and deception all seem to be present in one form or another during the period of January 2002 - March 2004.

Evidence of election tampering is present in the Alderwood Voter Registry. It appears that at least two different individuals have tampered with the registry. Both made mistakes which can be used to identify which records are fraudulent.

The first individual input dates of birth in an incorrect format; all records inserted by this individual have a date of birth of 8-Apr-34, and a registration date from one to four years after this date. These records also reuse the same set of first names several times. There are 73 records of this type inserted into the registry.

The second set of records was properly formatted, but the registration dates did not seem to correspond to the dates of birth listed. In the state of Washington, an individual must be 18 by the date of the next election to register to vote. Many of the records contained registration dates which preceded the date of birth listed. Other records contained registration dates which were well before (>10 years) the date on which the individual could conceivably register.

With the exception of the aforementioned records, the majority of the records indicate that the registered individual registered when they were between the ages of 17 and 40. There was one exception: Laurel Sulfate was registered on 8/1/2002, at the age of 16. This should not have been possible, since mid-term elections took place on Nov 5, 2002, before Sulfate was eligible to vote. This leads me to believe that Laurel Sulfate was somehow connected to the first instance of election tampering, since the style of her record best matched the style of that saboteur.

Someone in the Alderwood government seems to be involved with an entity in Switzerland, as well. The City Hall conference room records show calls to entities in Berne and Kriens on several occasions in January of 2002. Bruce Rinz has his own phone line as City Attorney, so it is not likely that he made these calls, and there is no indication of activity involving the Swiss development company this early in the timeline, so it is more likely that these calls were related to the founding of Boynton Labs.

Bruce Rinz also seems to be involved in something not entirely aboveboard. There is no listing in the voter registry for a Bruce Rinz, which seems strange for a public official. There is, however, a listing for Augustus Rinz. The individual serving as Interim City Manager is referred to in several articles as Bruce Rinz, and later as Mike Rinz. It seems plausible that Augustus Rinz is using an alias for some reason. If the date of birth reported in the voter registry is correct, Rinz would be 49 years old, which is probably in the correct age range for someone who is a lawyer and active in city government.

The voter registry indicates that Augustus Rinz was born in Switzerland. This may support a connection between Rinz and an entity in Switzerland. There are several entities in Switzerland which show up in the data set: a Swiss developer which may be interested in developing the land purchased from the Monson family, the entities contacted from the City Hall conference room in January of 2002, and the VonRyker Institute AG.

However, Rinz registered to vote at the age of 20, which indicates that he was a US citizen at this time. In order to become a naturalized citizen, Rinz would have had to reside in the US for at least 5 years prior to his application for naturalization. To establish the required residency period, Rinz would have needed to leave Switzerland on or before 1/7/72, which seems to reduce the probability of his connection to entities in Switzerland. This does, however, depend on the accuracy of the voter registry, which has already been called into question.

Rinz also seems to be financially connected to the Washen Foundation, an organization which provided a great deal of the funding for Boynton Labs. There seems to be enough evidence that Rinz is involved in something to warrant further investigation.

Hypothesis: Boynton Labs is engaged in unethical conduct.

Both Dr. Alejandro VonRyker and Dr. Delwin Sanderson, chief scientists at Boynton Labs, were involved in the VonRyker Institute AG, which was closed due to violations of bioethics, fraud, and misuse of research funds. The work done at the VonRyker Institute was very similar to that done at Boynton Labs. It may be that the same kind of unethical behavior that went on at the VonRyker Institute continues at Boynton Labs.

VonRyker faked two papers on synthetic transmissible prion diseases while director of the VonRyker Institute, and a news story from the Alderwood Daily News reports that Boynton Labs is continuing research in this area. It is possible that VonRyker and Sanderson are simply using Dr. Boynton, who is a reputable scientist, as a front for continuing unethical activities. The phone calls made from the City Hall conference room occurred immediately before the opening of Boynton Labs was announced and shortly after the VonRyker Institute was closed. If further research into these phone calls indicates that they support a link between Alderwood and the VonRyker Insitute, this may indicate that it was VonRyker who is truly in control of Boyton Labs.

The highly publicized bovine spongiform encephalitis cases found in the Alderwood area may have provided VonRyker and Sanderson an opportunity to open a lab where they could continue their research without arousing suspicion.

Laurel Sulfate’s vacation to Switzerland also seems too convenient to be coincidence. It definitely appears that there is a continuing connection between Boynton Labs and some entity in Switzerland.

Hypothesis: Boynton Labs is connected to the Alderwood city government.

There are clearly several personal connections between individuals in the Alderwood government and individuals employed by Boynton Labs, but the Alderwood city government may also be granting Boynton Labs political favors in an attempt to jump-start the local economy. Several people may also have personal reasons for wanting Boynton Labs to be successful.

There are rumors that Bruce Rinz is heavily invested in the Washen Foundation, which is a major stockholder in Boynton Labs. He may be using his positions as City Attorney and City Manager to influence city politics to support Boynton Labs.

Laurel Sulfate appears to be linked to election tampering, as well as being romantically linked to John Torch and being connected to both the Alderwood city government and Boynton Labs.

Laurel Sulfate is listed in the voter registry with a registration date of 8/1/2002, a full year before she would be eligible to register. There are over 150 other entries in the voter registry which give registration dates before the registered individual would be eligible. Sulfate worked as the Mayor Rex Luthor’s aide in 2002, so she may have had an opportunity to tamper with the voter registry at this time.

Photos from April 2004 show Sulfate with John Torch. These photos seem to indicate that some relationship exists between Torch and Sulfate. Since Torch was active as a councilman at this time, their relationship seems to indicate that Sulfate might be able to influence city politics. Since Torch and Sulfate worked together when Sulfate was the Mayor’s aide in 2002, their relationship may go back as far as 2002.

Beginning in September 2002, Sulfate is linked with Boynton Labs. On September 15^th, 2002, she is listed as executive assistant to Dr. Philip, and by September of the next year, she is listed as spokesperson for Boyton Labs. Her rapid advancement, combined with her connections in the Alderwood city government, may indicate that she is influencing the political situation in Alderwood to benefit Boynton Labs. Further investigation of Sulfate is probably wise.

5. VISUALS and Description of ANALYTIC PROCESS

The overarching strategy I applied to the Alderwood data set was to identify key players, pick out the events each of these key players was involved in, and discover the relationships between each of the players based upon their interaction.

My initial strategy in analyzing the data set was to perform broad searches to identify key players. DECIDE provides local search capabilities through integration with Windows Desktop Search, so I began by searching the provided news articles for general terms provided in the “Analysts background” document. Initial searches for “Alderwood Scientist” and “Boynton” returned a fairly long list of documents, which I examined using DECIDE’s Document Preview function.

Figure 1: DECIDE Desktop Search and Document Preview

From the documents returned in my initial searches, I created a list of keywords and names for further search. I maintained a mindmap in FreeMind to visually keep track of connections as I discovered them. My initial goal was to determine key players, key events, and important locations contained within the data set, to facilitate deep searches later in my analysis.

My initial searches revealed that some of the key individuals in the Alderwood city government were Rex Luthor, John Torch, Bruce Rinz, George Greenway, and Robert Rockford. There were a few names which were often seen in connection with Boynton Labs: Dr. Philip Boynton, Laurel Sulfate, Dr. Delwin Sanderson, and Dr. Alejandro VonRyker.

Figure 2: FreeMind mindmap diagramming connections between actors and events

Based on deep searches beginning with the names listed above, I began to create evidence items in DECIDE to capture information about each individual. An evidence item in DECIDE represents a distinct datum which is evaluated for credibility based on a set of attributes describing the reliability of the source, the plausibility of the information represented in the item, and the quality of the information.

Initially, I created evidence items detailing the connection between Laurel Sulfate and John Torch; the connection between Bruce Rinz, the Washen Foundation, and Boynton Labs; the phone calls made during January of 2002; and the issues of fraud and ethical misconduct associated with the VonRyker Institute AG.

I attempted to investigate links between individuals by searching names in combination with each other, but this turned out to be less helpful than I expected. The set of individuals I was interested in tended to show up in most of the same documents, so searching for “Boynton AND Sulfate” returned almost as many documents as searching for “Sulfate” alone.

Figure 3: Overview screen in DECIDE showing hypothesis with evidence added.

Counterclockwise from top right: Boynton Labs hypothesis, circular layout; Political Corruption hypothesis, radial tree layout; Alderwood connected to Boynton Labs hypothesis, tree layout

After creating my initial collection of evidence, I went back to the original background document and reviewed the information which prompted the investigation of Alderwood. The initial background information mentioned that the influx of high-tech talent into Alderwood was supported by the “high-rolling bigboys at City Hall,” so based on this premise, I created three hypotheses: political corruption is present in the Alderwood city government, Boynton Labs is involved in unethical conduct, and that there exists some degree of connection between the Alderwood city government and Boynton Labs.

The conference room phone log and voter registry came up repeatedly in my initial searches, but I hadn’t explored them very thoroughly at this point in my analysis, so I decided to investigate them further. Several of the news stories in the data set had mentioned absentee ballots, and I reasoned that if I were attempting to tamper with an election, the safest way to do it would be to insert false records into the voter registry and then vote via absentee ballot.

Assuming that tampering of this kind was going on, I examined the records in the voter registry in Microsoft Excel. I sorted each column individually and examined the high and low values. This revealed that the registry contained improperly formatted dates in the DOB column. I examined those records and noticed that the registration dates listed did not match either. In order to more easily examine the difference between date of birth and registration date, I wrote a function in a spare column to calculate the difference in years. Sorting on this column revealed that there were other records which were invalid. It also revealed that Laurel Sulfate registered to vote a full year before she was eligible. I created several evidence items to capture the information I had discovered in the voter registry.

Glancing over the conference room phone log, several calls jumped out immediately. The majority of the calls were made within the 509 area code, but five calls were made that required a different country code, and five calls were made to different area codes. Some quick internet research revealed that the international calls were to Berne and Kriens, Switzerland. The long-distance calls were to law firms in Seattle, and to a number in Toronto.

DECIDE’s overview screen provides several ways of examining arguments based on statistics, structure, and evidential content. I used the comparison view to examine the evidence in my collection, and noticed a spike of activity in early 2002 (the highlighted section of the image below).

Figure 4: Overview screen in DECIDE showing a breakdown of evidence by date and containing argument

The color of a particular bar in the graph is a visual indicator of the age of the evidence items represented relative to the ages of all evidence items displayed in the current view.

The highlighted area shows a particularly active period: in this case, the period when the phone calls listed in the conference room phone log were made.

Further examination of the evidence from that period revealed that most of the activity was due to the phone calls recorded in the conference room phone log, but there were also several events related to bovine spongiform encephalitis, which suggested that there might be a connection. The closing of the VonRyker Institute and the founding of Boynton Labs occurred immediately before and after this spike of activity. This information, as well as the fact that the VonRyker Institute was located in Switzerland and the connection between Dr. VonRyker and Boynton Labs, suggested a connection between the VonRyker Institute, Boynton Labs, and the Alderwood city government.

Figure 5: Overview screen showing final hypotheses.

Sub-hypotheses connected by red dashed lines indicate alternate hypotheses which could provide evidence which does not support (and may refute) the parent hypothesis.

After reviewing my hypotheses, I revised them to include possible alternate hypotheses. These alternate hypotheses provide an explicit definition of possible alternate explanations which would not support the root hypothesis, and provide structure for intelligence collection tasking. The data set provided little support for the alternate hypotheses I devised, but by enumerating alternate possibilities, I aimed to eliminate getting locked into one view of the data that looked promising.

I examined my final hypotheses using DECIDE’s Analyzer screen, which uses Heuer’s method of Analysis of Competing Hypotheses to evaluate the plausibility of a given hypothesis based on evidence. The image below shows that each hypothesis has a high probability of being valid, given the current set of evidence.

Figure 6: Analyzer screen in DECIDE.

This screen evaluates hypotheses using Heuer’s method for Analysis of Competing Hypotheses.

The final step in this analysis would be additional evidence collection to establish the plausibility of alternate hypotheses. At this point, the hypotheses listed above seem plausible, but additional investigation is required to either confirm the primary hypotheses or provide support for alternate hypotheses.

Russ Lankenau and Steve Eick accepting our award.

Part of the award was a session with an analysts trying to solve a real problem.

We placed 2nd on the Visual Analytics Science and Technology contest!  The contest was sponsored as part of the IEEE Vis 2006 conference.  DECIDE is one of our core products used in hypothesis analysis.

IEEE VAST 2006 CONTEST
-----------
A tale of Alderwood

First place, Corporate Category
Applied Analysis with nSpace and GeoTime
Pascale Proulx, Lynn Chien, Adam Bodnar, Kaleb Ruch and William Wright, Oculus Info Inc.

Second Place , Corporate Category
DECIDE™
Russell A. Lankenau, M. Andrew Eick, Alexander Decherd, Maxim Khailo, Phil Paris, and Jesse Fugitt, SSS Research, Inc.

First Place , Student Category 
Visualizing Relationships in a Diverse Data Collection
Summer Adams and Kanupriya Singhal, Georgia Institute of Technology

Source: IEEE VAST 2006 CONTEST (cached)

http://blog.thincsoft.com/PermaLink,guid,c48d714a-7f1f-4d0e-ba1f-37be52d5da7e.aspx http://blog.thincsoft.com/2006/12/28/Visual+Analytics+Science+And+Technology+Contest.aspx Thu, 28 Dec 2006 00:25:35 GMT <p> We placed 2nd on the Visual Analytics Science and Technology contest!&nbsp; The contest was sponsored as part of the <a href="http://vis.computer.org/vis2006/index.html">IEEE Vis 2006 conference.</a>&nbsp; DECIDE is one of our core products used in hypothesis analysis. </p> <blockquote> <h3> </h3> <h3><a href="http://blog.thincsoft.com/content/binary/WindowsLiveWriter/VisualAnalyticsScienceandTechnology_123F1/IEEE_Logo7.png" atomicselection="true"><img style="border-top-width: 0px; border-left-width: 0px; border-bottom-width: 0px; margin: 5px; border-right-width: 0px" height="145" src="http://blog.thincsoft.com/content/binary/WindowsLiveWriter/VisualAnalyticsScienceandTechnology_123F1/IEEE_Logo_thumb5.png" width="350" align="right" border="0"></a>IEEE VAST 2006 CONTEST<br> -----------<br> A tale of Alderwood </h3> <p> <b>First place, Corporate Category</b> <br> Applied Analysis with nSpace and GeoTime<br> <i>Pascale Proulx, Lynn Chien, Adam Bodnar, Kaleb Ruch and&nbsp;William Wright, </i>Oculus Info Inc. <p> <b>Second Place</b><b>, Corporate Category</b> <br> DECIDE™<br> <i>Russell A. Lankenau, M. Andrew Eick, Alexander Decherd, Maxim Khailo, Phil Paris, and Jesse Fugitt, </i>SSS Research, Inc<i>.</i> <p> <i></i> <p> <b>First Place</b><b>, Student Category&nbsp; <br> </b>Visualizing Relationships in a Diverse Data Collection<i> <br> Summer Adams and Kanupriya Singhal, </i>Georgia Institute of Technology </p> </blockquote> <p> Source: <a href="http://www.cs.umd.edu/hcil/VASTcontest06/">IEEE VAST 2006 CONTEST</a>&nbsp;(<a href="http://www.andyeick.com/_blogMedia/2006_IEEE.mht">cached</a>)<a href="http://www.cs.umd.edu/hcil/VASTcontest06/"></a> </p> <img width="0" height="0" src="http://blog.thincsoft.com/aggbug.ashx?id=c48d714a-7f1f-4d0e-ba1f-37be52d5da7e" /> http://blog.thincsoft.com/CommentView,guid,c48d714a-7f1f-4d0e-ba1f-37be52d5da7e.aspx News