This proposal responds to the August 1999 Solicitation for Science and Technology by the National Institute of Justice. The Tucson Police Department, Phoenix Police Department, and the University of Arizona Artificial Intelligence Group propose to partner in the continued research and development of current state of the art and near term future database and Intranet technologies to enhance inter-agency sharing and analysis of information among criminal justice agencies. This partnership will build upon the COPLINK application that was developed in a two-year effort that began this partnership between cutting-edge information systems research and law enforcement. The COPLINK prototype has inspired a consortium of agencies in the state of Arizona committed to using this valuable tool to make a historical advancement in police information sharing. If this proposal is funded, the first multi-agency COPLINK Intranet will be developed and deployed, linking the Phoenix Valley and Tucson, and further development of the application will occur, including expansion of the database and interface to accommodate a richer array of available data. Concurrent with this deployment will be the continued research and development of cutting edge analysis and knowledge management tools. The intended areas of research in textual analysis techniques include linguistic analysis, semantic interoperability, semantic parsing, and visualization and summarization techniques. The group also intends to pursue development of law enforcement intelligent agents (spiders) for both the COPLINK Database/Concept Space applications and for Internet searching. Continued research in distributed database and system design will also continue to refine COPLINK for greater interoperability and platform and application independence.
Dr. Hsinchun Chen is head of the University of Arizona/MIS Artificial Intelligence Group. Dr. Chen has been heavily involved in fostering digital library and knowledge management research in the US and internationally. He is also a Visiting Senior Research Scientist at National Center for Supercomputing Applications (NCSA). Dr. Chen's work has been recognized by major US corporations and been awarded numerous industry awards.
The City of
Tucson Police Department (TPD) requests funding to deploy and develop the
COPLINK Database/Concept Space distributed system for TPD and Phoenix Police
Department (PPD), and to continue the development of cutting edge tools to
enhance these applications and provide advanced knowledge management techniques
for law enforcement. This funding will establish the COPLINK Center for Law
Enforcement Information Sharing and Knowledge Management.
The Tucson Police Department (TPD) is committed to developing innovative and cost effective methods to better serve the community. TPD’s mission statement charges the agency to "To serve the public by furthering the partnership with our community, to protect life and property, prevent crime and resolve problems". TPD is deeply committed to a philosophy of community policing, which it officially implemented in 1995 (see Appendix A). Within the last year, the Tucson Police Department has made improvements to its information technology infrastructure that will help the agency to meet this challenge. TPD has provided important community information to the community it serves through its web site and public access kiosks. The unique partnership that the agency has developed with the University of Arizona exemplifies this collaboration with its community partners, and TPD’s commitment to providing better service to the public.
TPD is also committed to providing better access to information to officers and investigators to improve efficiency and effectiveness in its crime fighting efforts and protection of the public. TPD is in the process of upgrading its central computing systems to accommodate richer sources and uses of information including voice and multimedia capabilities. TPD has also invested personnel, equipment and additional $245,000.00 in funding to implement the COPLINK Database and COPLINK Concept Space applications. TPD is committed to using this application to enhance its information sharing capabilities within the department and with outside agencies.
Some ongoing problems for law enforcement agencies are: lack of integration and lack of secured, cost-effective, and timely access to criminal justice information. This isolation of criminal information databases has severely impacted agencies’ attempts to use information technologies to effectively prevent crime and provide safety to their communities. Most mid and large sized police agencies have such systems which provide access to information by their own personnel, but lack any efficient manner by which to provide that information to other agencies. Government entities, the criminal justice community included, have commonly lagged behind the business community in information systems practices in general. This delay by government entities in implementing new technologies can be attributable to several factors. Government entities rely on funding that is typically short term or sporadic in nature. Government entities are also not driven by competition or innovation. Research and development tends not to be funded locally; states, counties and municipalities must demonstrate immediate benefit to their communities from resources allocated to specific projects. Most research and development resources come from federal agency funding. The oftentimes piece-meal, short-term approach to information technology adaptation has caused government entities to exhibit fragmented information management and sharing.
Criminals show no regard for jurisdictional boundaries and in fact take advantage of the lack of communication across jurisdictions. This problem places criminals at an advantage.
For example, in the State of Arizona, the major population centers of Tucson and Phoenix are located roughly 100 miles apart. Gang members from Phoenix commonly travel to Tucson to commit crimes such as robberies and car thefts, then return to Phoenix. Investigators in Tucson can only access case report information from Tucson, unless they can identify a specific case report to request from the Phoenix Police Department. Thus, valuable information that could provide investigative leads and apprehension of suspects is not shared, and criminals continue to victimize citizens.
Crime and police report information is rapidly migrating from paper records to automated records management databases. Federal standards initiatives such as the National Incident Based Reporting System, or NIBRS, (US Department of Justice, 1998) are attempting to provide reporting standards to police agencies to facilitate future reporting and information sharing among agencies as these electronic reporting systems become more widespread. Database technology traditionally plays an important role in the management of information for police departments. Previous research has detailed the use of database technology to organize information in a form that can be easily searched by officers and other employees in a police department (Hoogeveen & Van der Meer, 1994; Lewis, 1993; Lingerfelt, 1997; Miller, 1996; Schellenberg, 1997; Wilcox, 1997a). The use of relational database systems for crime-specific cases, such as gang-related incidents, and serious crimes, such as homicide, aggravated assault, and sexual crimes, has proved to be highly effective (Fazlollahi & Gordon, 1993; Pliant, 1996; Wilcox, 1997b). The use of databases in these criminal areas is often targeted because it allows for a manageable amount of information to be entered into the database and can also combine information that may normally exist in neighboring police districts.
A distributed database system “is an environment in which data in two or more database instances is accessible as though this data were in a single instance” (Dye, 1999). This implies communication between two or more databases physically separated on separate server nodes. Implicit in this concept of a distributed database system is an underlying network through which the physically separated databases communicate.
With the increasing emphasis on “content” in large web-based applications, distributed architecture is becoming more popular, as web-enabled database systems move away from a client-server (2-tier) modality to a three-tier architecture. Typical of a three-tier architecture are a “thin” web-based client, a web application server, and a database server. This is increasingly becoming the norm in web-enabling distributed database systems (Yang 1998; Owens, 1999). The COPLINK system implemented with the current NIJ funding is an implementation of such architecture, with a graphical Java client interface, an Oracle Web Application Server (WAS), and an Oracle database (See Appendix C).
In this era of the Internet and distributed multimedia computing, new and emerging classes of information technologies have swept into all fields of business, industry and government. As information technologies and applications become more overwhelming, pressing, and diverse, information technology problems have become even more urgent. Information overload, a result of the ease of information creation and transmission via Internet, the WWW, and organizational data sources has become more evident in people's lives (Blair & Maron, 1985) (Chen, Martinez, et al., 1998). This phenomenon is nowhere more evident than in the field of government, specifically in criminal justice information systems. Federal, state, and local criminal justice institutions have long held or accessed vast repositories of information. With the explosive growth in digital information and access within government agencies, information overload becomes increasingly prominent. Significant variations of database formats, structures, and content, and the richness of information media (text, audio, and video), also have created different information interoperability problems – structural interoperability and media interoperability (Paepcke et al., 1996) (Lesk , 1997).
As sharing of police records information becomes more commonplace, the problems of knowledge management that exist in business, science, industry, and other facets of government will only become more prevalent in law enforcement. Ease of capture, retrieval, and access of law enforcement data is increasing thus information overload increases proportionately. Large textual collections exist in police records in the form of report narratives and statements that are ripe for the development of textual mining and linguistic analysis applications.
In law enforcement, knowledge about criminal activities or specific groups and individuals tends to be learned by officers who work in specific geographic areas. The information is often lost with personnel changes and has to be reacquired by new officers. Valuable information is often stored in police databases, but the tools necessary to retrieve and assemble it do not yet exist or are insufficiently refined for the specific task. Solving problems by analyzing and generalizing current criminal records is a function of the daily routine of many crime analysts and detectives. Potent intelligence tools can be useful in the analysis of available criminal records and aid in the investigation of current cases by alleviating the crime analysts’ information overload and reducing information search time.
Recognizing this severe information technology problem in the current information society, law enforcement agencies across the United States have begun to explore and adopt innovative information and knowledge management technologies to aid in the sharing of criminal information. Such technologies can allow multiple agencies to share time-critical, life-saving information and could potentially serve as intelligence tools to combat criminal activity by aiding in case investigation or even predicting criminal activity.
Several new intelligence analysis tools for law enforcement have begun to emerge. Some systems use neural networks to solve problems by developing associations between information objects and are trained to solve problems by comparing known objects with unknown objects. Some applications utilize visualization and time analysis to examine information. For example, the Timeline Analysis System (TAS) can help analysts visually examine large amounts of information by illustrating cause-and-effect relationships. This system graphically depicts relationships found in the data, resulting in trends or patterns (Pliant, 1996). Expert systems that employ rule-based information have also been developed to assist in knowledge-intensive activities (Bowen, 1994; Brahan, 1998). These systems attempt to aid in information retrieval by drawing upon human heuristics or rules and procedures to investigate tasks.
Based on the first COPLINK law enforcement project funded by the National Institute of Justice, this project aims to continue the development of the COPLINK application and technologies for breakthroughs in information sharing in law enforcement by establishing the COPLINK Center for Law Enforcement Information Sharing and Knowledge Management.
The University of Arizona/MIS Artificial Intelligence Group (AI Group) has done extensive NSF-funded research in digital library and knowledge management. The technologies they have developed in linguistic analysis, semantic parsing, information visualization, and intelligent agents will form the basis for continued development of these techniques for law enforcement.
The Phoenix Metro Area when combined with the City of Tucson Metropolitan Area contains 76% of the State's total population. Presently, there is no crime analysis data sharing between these two major population centers. Phoenix and Tucson are only 100 miles apart and it is not uncommon for the same criminals to work both areas. The University of Arizona MIS/Artificial Intelligence Group (AI Group), headed by Dr. Hsinchun Chen, will continue to collaborate with TPD and PPD to achieve research goals related to law enforcement information and knowledge management.
The City of Tucson incorporates approximately 194 square miles and has a population of approximately 470,000 people within the city limits, approximately 750,000 people in the Tucson Metropolitan area. The Tucson Police Department (TPD) is comprised of approximately 900 commissioned police personnel and nearly 300 civilian support and supervisory personnel. A research partnership is ideal for an agency the size of the Tucson Police Department (TPD). While TPD experiences “big city” problems, it is still small enough to implement, oversee, and measure a variety of research issues.
TPD stores a portion of each incident report generated in a central Records Management System (RMS). The Tucson COPLINK Database currently integrates three data sources: The TPD RMS, the TPD Gang Unit database, and the TPD Video Mug Shot database. This integrated database currently contains approximately 1.4 million incident records and will continue to grow as the Tucson Police Department adds additional records to its RMS. TPD is committed to fully deploying the COPLINK Database and COPLINK Concept Space within the agency. In fact, TPD has now allocated an additional $245,000 to the COPLINK project to fully operationalize Concept Space and to develop greater integration and ease of use between the COPLINK Database and COPLINK Concept Space. The Department has also allocated approximately 2.5 persons to staff the project during this TPD funded effort. TPD will continue to provide domain expertise to the project.
The Tucson Police Department is scheduled to implement an automated field reporting system over the next year and plans to include the narrative portion of the reports within the automated record. This unstructured text will then be integrated into the Tucson COPLINK node. This reporting system will provide researchers with an additional and growing collection of report narratives.
The Phoenix Police Department employs more than 3,500 officers and civilian support personnel. The Department is tasked with serving and protecting a population of over 1,400,000 and patrols an area of just under 485 square miles.
The key operational philosophy of the Phoenix Police Department is community based policing. Working closely with other valley agencies, city departments, and citizens, the Department has forged a community based policing partnerships that are productive for all concerned, and have helped it earn a national reputation for excellence.
In an effort to continue providing the best services for our community, the Department is focused on the use of technology to further increase its productivity. The Department recently approved adding eight new technical positions bringing the authorized staff count in its Computer Services Bureau to 60. The Computer Services Bureau has recently implemented a new high‑speed data communications network linking its 24 facilities. A Crime Analysis System complete with GIS spatial analysis capabilities has been created to enhance crime solving and resource allocation activities. The Department's extensive records management capabilities have been further enhanced by moving to a much faster hardware configuration.
The Department believes that further significant crime analysis advancements are dependent upon sharing case and field interrogation information between agencies. The Phoenix metropolitan area consists of the cities of Phoenix, Glendale, Tempe, Mesa, Scottsdale, Gilbert, Chandler and Paradise Valley. Each city has its own police department and each department has its own unique records management system. Each system is used to collect information on individual crimes, field interrogations and arrests. This data is used for internal case management and crime analysis within each agency. Data used to support crime analysis and to develop case leads is not shared between agencies. Unfortunately, each police agency's ability to obtain data needed to solve crimes and apprehend criminals ends at its city limits.
The Department believes that COPLINK is the solution to its data sharing needs and is firmly committed to working with the Tucson Police Department and the University of Arizona to implement COPLINK throughout the state. The Department has received City Council approval to spend up to $460,000.00 to purchase the computer equipment and software needed to implement a COPLINK enterprise server for the Phoenix metro area, and to interface with the COPLINK system in Tucson. The Department is also committing an equivalent of 2.5 highly skilled computer professionals for the duration of the project. Agency participation in the Phoenix metro area is presently committed at 100% (See Appendix E).
The Department believes that, to be successful, it will need assistance from the University of Arizona in the form of training and implementation support. The Department also believes that further planned COPLINK enhancements are significant in terms of their ability to improve our crime solving capabilities.
The Phoenix Police Department believes that the potential benefits that will come from successful COPLINK implementations in Phoenix and Tucson will lead to other COPLINK installations in other population centers across the United States. The Department hopes that the NIJ will give favorable consideration to our collective request funding to continue this worthwhile project.
The State of Arizona has recently implemented a statewide mug shot database. As Phoenix begins to implement its COPLINK node, the COPLINK team will begin to evaluate the state mug shot system for integration with COPLINK. Once the majority of Arizona police agencies convert to the state system over the next few months, it will store approximately 500,000 mug shots and driver’s license photographs.
The Arizona Department of Public Safety (AZ DPS) has given
approval to the COPLINK project participating agencies to use the Border States
Network infrastructure. This
infrastructure was funded to enhance sharing of narcotics trafficking and
intelligence information in California, Arizona, New Mexico and Texas. This network will support access to remote
agency information.
The University of Arizona, Management Information Systems Department Artificial Intelligence Group, headed by Dr. Hsinchun Chen, will be the research partner for the proposed project. Dr. Chen received the NSF Research Initiation Award in 1992 and the HICSS Conference Best Paper Award in 1994. Dr. Chen’s work also has been recognized by major US corporations and been awarded numerous industry awards for his contribution to IT education and research. In 1995 and 1996, he received the AT&T Foundation Award in Science and Engineering. In 1998 he received the SAP Award in Research/Applications and became the Karl Eller Center Honored Entrepreneurial Fellow. In 1999, Dr. Chen received the McClelland Endowed Professorship and the Andersen Consulting Professor of the Year Award.
Dr. Chen has been heavily involved in fostering digital library research in the US and internationally. He was a PI of the NSF-funded Digital Library Initiative-1 project (1994-1998) and he also recently received another major NSF award (1999-2003) from the new Digital Library Initiative-2 program. Dr. Chen was the guest editor of digital library special issues in IEEE Computer (May 1996 and February 1999) and Journal of the American Society for Information Sciences (1999, forthcoming). He also helped organize the Asia digital library research community and chaired the First Asia Digital Library Workshop, held in Hong Kong in August 1998. He is the conference chair of the Second Asia Digital Library Workshop, to be held in Taipei in November 1999.
Dr. Chen has frequently served as a panel member and/or workshop organizer for major NSF and DARPA research programs. He has helped set directions for several major US initiatives including: the Digital Library Initiative (DLI), the Knowledge and Distributed Intelligence Initiative (KDI), and the Integrated Graduate Education and Research Training (IGERT) program.
The Project Manager for the University of Arizona under Dr. Chen will be Dr. Homa Atabakhsh. Dr. Atabakhsh received her B.Sc., M.Sc. (1984). and Ph.D.(1987) in Computer Science from the University of Toulouse, France. Her Ph.D. research was in the area of Knowledge based systems and their application in modeling of flexible manufacturing systems. From 1988 to 1989, she was an Assistant Professor in Toulouse teaching Computer Science courses such as algorithmic programming, operating systems, C programming language, and Unix. In 1989, she was employed by the National Research Council of Canada as an assistant Research Officer, later promoted to associate Research Officer. During this time she worked in areas such as: object oriented design and programming, graphical user interface design and programming, knowledge-based systems, integration of rule-based systems and object oriented systems, discrete event simulation, and heuristic base scheduling. Since Jan. 1999 she has been adjunct lecturer at the department of Management Information Systems (M.I.S.) of the University of Arizona teaching object oriented design and programming, graphical user interface design, Visual C++ and Java programming languages. Since Sept. 1999, she has been Principal Research Specialist at M.I.S. working as project manager for the Coplink project.
The University of Arizona, located in Tucson, is a first-tier research institution and has collaborated extensively with local and state governments and law enforcement agencies over the years. The University has provided education training and research supports for local communities and law enforcement agencies.
The Management Information Systems (MIS) Programs at the University of Arizona are designed to apply relevant computer technology, quantitative techniques, and administrative skills to the information technology problems of organizations. The Department of MIS offers degrees at the B.S., M.S., M.B.A. and Ph.D. levels. It has been ranked from fifth to third nationally by US News & World report for the past eight years. This ranking is determined by the vote of more than 270 MBA directors and business deans of institutions accredited by the American Assembly of Collegiate School of Business (AACSB). The department was rated second in the Database, Fall 1992 rankings of Top Institutional Representation in MIS literature over the periods 1982 to 1991. The department’s reputation has resulted in more than $30,000,000 research and instructional support over the years. Funding and equipment grants have come from major corporations and government agencies.
The UA/MIS Artificial Intelligence Group consists of 10 Ph.D.-level researchers and 25 research scientists and assistants. It specializes in database integration, digital libraries, knowledge discovery, Internet/Intranet technologies, and intelligent information retrieval. It has received more than $15 million in funding from various government agencies including National Science Foundation (NSF), Advanced Research Projects Agency (ARPA), National Aeronautics and Space Administration (NASA), and National Institutes of Health (NIH). Among its ongoing large-scale database and digital library projects are:
1.
DARPA-funded Information
Management project on “The Interspace Prototype: An Analysis Environment based
on Scalable Semantics,” 1997-2000 ($3.2M): The proposed analysis environment for Geographic
Information Systems (GIS) will be a “semantic middleware” for scalable
distributed services, which consist of a hierarchy of statistical and pattern
analysis algorithms that are valid across the range of object types (e.g.,
texts and images) and subject domains.
2.
NSF/NASA/ARPA-funded
“Digital Library Initiative” (DLI Phases 1 and 2) project, "Building the
Interspace," 1994-2002 ($5M): The project goal is to
develop a large-scale testbed for building next-generation digital libraries
for the National Information Infrastructure (NII). Semantic retrieval and indexing and vocabulary switching
techniques have been developed and deployed on the testbed. The testbed collections are mainly in the
engineering domains, to be contributed by major engineering societies and
publishers.
3.
NIH-funded ``Semantic
Retrieval for Medical Informatics'' project, 1996-1999 ($500K): This project aims to create concept spaces and
category maps for medical literature, including CancerLit and Toxline
databases. The resulting machine-generated
knowledge structures have been integrated with the NLM's Unified Medical
Language System (UMLS) to enhance medical information retrieval and knowledge
sharing.
4.
National Center for
Supercomputing Applications (NCSA) High Performance Computing Resources Grant
“Information Analysis and Knowledge Management for Digital Libraries'' project,
1995-1999: This project aims to develop scalable,
high-performance computing techniques to support concept-based knowledge
management for large-scale multimedia (text and image) digital libraries.
5.
NSF-funded ``An Intelligent
CSCW Workbench: Personalized Analysis and Visualization'' project, 1998-2001
($300K): This
project aims to create an intelligent workbench to explore research issues related to coordinated work.
Advanced categorization and visualization (1D, 2D, 3D) techniques for
collaborative computing are under development.
6. NSF-funded ``Internet Categorization and Search'' project, 1995-1998 ($210K): This research, which is grounded on automatic textual analysis of Internet documents (homepages), attempts to address the Internet/Intranet search problem by automatically categorizing the content of Internet documents and providing semantic search capabilities based on inductive machine learning algorithms and agents.
Research findings from the Artificial Intelligence Group have been featured in Science, Government Computer News, NCSA Access Magazine, WEBster, HPCWire, TECH Beat, and New York Times.
·
“It’s
called a Web-based intuitive integrated interface. But in layman’s terms it’s
called Coplink. What it will do is help put an end to a serious problem faced
by law enforcement every day...the inability to exchange information about
criminal cases across jurisdiction.” -- “Coplink: Database Detective,” TECH Beat,
Summer 1999.
·
“In
a different approach, the artificial intelligence lab in the management
information systems department of the University of Arizona at Tucson drew a
map of more than 100,000 entertainment Web sites...The map looks something like
a jigsaw puzzle with colored pieces. Users can select categories like movies or
comics, or terms like ‘love’ or ‘beer’, and the map will adjust.” – “Beyond
Geography: Mapping The Unknowns of Cyberspace,” New York Times, Technology
Section cover article, September 30, 1999.
·
"COPLINK
intranet [designed by the AI Group] will bring Arizona crime fighters to the
data they need. In fact, there's no reason it couldn't connect all the police
departments nationwide." -- Government Computer News, January, 1998.
·
"Towards
Concept Search...Concept spaces and vocabulary switching [developed by Dr.
Hsinchun Chen] will need to be part of the fundamental infrastructure if
digital libraries are to support correlations between information sources at
all these levels." -- "Bring Search to the Net," Science, 17
January, 1997 Cover Article
·
"Now,
with little fanfare and no sonic boom, Schatz and Hsinchun Chen of the
University of Arizona have opened what they claim is the `first crack in the
semantic barrier.' What they've done is lay the groundwork for a system that
would provide a user with key words needed to search for information across
fields." -- "Computation Cracks `Semantic Barriers' Between
Databases," Science, 7 June 1996.
· "A year ago, no one would have thought of information sciences as posing a supercomputer problem, and here it is, overnight, blossoming into one of the largest users of supercomputer time at NCSA. And, since the results of the computation will be useful to many of the faculty and students accessing the Illinois Digital Library testbed, the results of this work have wider applicability than any previous supercomputing application." -- Larry Smarr, Director of NCSA (National Center for Supercomputing Applications)
The COPLINK application has great promise in solving many information integration and access problems for law enforcement. The partnership between the Tucson Police Department, Phoenix Police Department and the University of Arizona provides a unique opportunity to blend law enforcement expertise and cutting edge information technologies. The established partnership between the University of Arizona Artificial Intelligence group and the Tucson Police Department has taken advantage of the first-tier research group’s proximity to a medium-sized police department that is moving ahead technologically. This successful partnership is now enhanced and expanded by a large city police department that is committed to implementing the technologies developed in Tucson and deploying, testing and further refining them on a larger scale. This partnership also provides the opportunity to evaluate and test network solutions and security for the transmission of sensitive law enforcement data over wide areas.
The Artificial Intelligence group will collaborate with TPD and the Phoenix Police Department in the following areas:
·
Creation
and development of Tucson and Phoenix COPLINK law enforcement databases and
concept spaces.
·
Development
of a secured, integrated, scalable, distributed COPLINK network for Arizona law
enforcement information sharing.
·
Field
studies and assessment of law enforcement information sharing needs and
solutions via COPLINK database and concept space technologies.
·
Research
and development in linguistics-based case (narrative) indexing and analysis.
·
Research
and development in dynamic, graphical case summarization and analysis based on
self-organizing map (SOM) neural networks.
· Development of law enforcement specific Internet agents/spiders for COPLINK Intranet information sharing (via information pushing) and Internet law enforcement information filtering.
The original COPLINK award allowed for the development and deployment of the COPLINK prototype. The vision for this prototype has always been that it spread to other agencies and be networked into a distributed information sharing system. Now, this application is poised to do just that. The COPLINK team has provided demonstrations of the prototype throughout the state. Numerous municipal, county, state, and federal agencies have expressed interest in COPLINK (See Appendix E).
The Tucson Police Department has allocated additional equipment and funding to fully deploy the application within the Tucson Police Department. Additionally, the Tucson Police Department is providing an additional $245,000.00 in funds as well as personnel and resources to fully deploy and operationalize the Concept Space application, and to develop greater integration and ease of use between the two applications.
The Phoenix Police Department is committed to deploy the first partner implementation of the COPLINK application. The Phoenix Police Department has already allocated personnel and equipment funding to bring up an initial COPLINK Phoenix prototype over the next 8 months.
As this second implementation of COPLINK occurs, development and refinement of the distributed COPLINK system is the next logical step. With the expansion of COPLINK into multi-agency information sharing, the opportunities for research and development in its analysis, knowledge retrieval and management capabilities also expand. The COPLINK team will take advantage of these opportunities to develop state-of –the-art analysis and retrieval techniques to enhance the COPLINK application and knowledge management techniques for law enforcement nationwide through the COPLINK Center for Law Enforcement Information Sharing and Knowledge Management.
The partnership has already received approval from the Border States Network project to use the new network infrastructure that has been implemented throughout the Southwest border states. This infrastructure gains vast opportunities for the COPLINK project to expand to neighboring states in the future. In addition, using this infrastructure leverages investments already made by federal programs to foster just such information sharing between police agencies.
Note: All Privacy Certificate Guidelines are currently adhered to by the COPLINK team and will continue to be adhered to in the proposed project. All law enforcement data will continue to be kept only at the Tucson Police Department, Phoenix Police Department, or at the secure COPLINK development facility at the University of Arizona. All members of the development team are subjected to a background questionnaire, fingerprinting, criminal history check, and Terminal Operator Certification Test in accordance with state and federal guidelines (See Appendix H).
TPD and PPD will now join with UA/MIS Artificial Intelligence Group in developing COPLINK database/Concept Space into a fully operational distributed, secured network for law enforcement information sharing. The partnership will collaborate as well in the continued development of future analysis and Internet/Intranet technologies for law enforcement.
The proposed project would be scheduled to conclude three years from the funding award date, during which TPD has budgeted 6,240 of (unfunded) personnel time to provide domain expertise and technical support to the Artificial Intelligence Group research scientists. The proposed research and development project addresses two major categories in the solicitation. The following areas of technological research and development work are proposed.
In 1997, the National Institute of Justice (NIJ) Office of Science and Technology funded a partnership between the Tucson Police Department and the University of Arizona Artificial Intelligence Group to address information accessibility problems common to law enforcement (Tucson Police Department, 1997b). The COPLINK application is specifically designed to share case and investigative information between law enforcement agencies in a secure Intranet. The integration of these records supports textual mining and analytical applications such as the law enforcement Concept Space (See Appendix C).
As stated earlier, the Phoenix Police Department is committed to implementing COPLINK in the Phoenix Valley, and has secured some funding for equipment and personnel for this endeavor. The COPLINK team has developed an initial system design and prototype that would allow multiple agencies with similar implementations of the COPLINK application to share information between agencies. This system design must be further developed to fully support a distributed implementation of the COPLINK Database and of the COPLINK Concept Space application.
The COPLINK team has made many design decisions based on the vision of many agencies sharing the COPLINK Database/Concept Space application. However, the group has always recognized that the first mirror implementation of COPLINK will necessitate some redesign. Since agencies capture virtually the same types of information (i.e. information about persons, locations, vehicles, incidents), the changes necessary to accommodate new agencies will drop drastically with each new implementation.
As the Phoenix Valley implements the COPLINK application, continued development and refinement of the application must occur to accommodate differences in the amount, types and representation of available data. As this refinement occurs, the team plans to evaluate additional information to be included in the database, interface, and search procedures. This evaluation will include the analysis of biometric information fields. The COPLINK team recognizes that strategies to uniquely identify persons across jurisdictional boundaries must be explored. These strategies most likely include the use of biometric indicators such as fingerprints, DNA, and other physical scans and scaling. The COPLINK team will evaluate some of these indicators. Viable candidates to include in searches will be tested and evaluated for addition to the database and search interface.
As changes are implemented to all levels of the system, the
COPLINK team will closely evaluate the initial prototype design. Now that a
prototype is in place, changes can be made based on the newest tools and
available technologies that will provide the greatest platform and application
independence. Refinements to the interface and web application server levels
may provide greater ease of development over varied ODBC-compliant
databases. A goal for the distributed
design is for application and platform independence of the underlying
databases.
Based on the initial design of a single COPLINK “node” for Tucson Police Department and the initial Phoenix COPLINK prototype, the COPLINK team will develop an expanded distributed system plan of multiple COPLINK “nodes” for an Arizona-wide law enforcement network (later to be expanded to other southwestern states). From a functionality standpoint, this would allow any law enforcement agency that implements a COPLINK node to share and access incident-base law enforcement information from any other participating agencies. The proposed system includes a “distributed” version of the COPLINK application, a modified set of queries that automatically query multiple COPLINK nodes, and the underlying system architecture to support a distributed COPLINK system. Significant database optimization and network security issues need to be studied carefully in the proposed distributed COPLINK project. The group will analyze existing connectivity and interoperability projects such as Florida’s CJ-Net (Florida Department of Law Enforcement, 1997), and the Colorado Justice Information Network (CJIN) to leverage their research and evaluation of user authentication techniques and secure network strategies, such as those using PKI (Public Key Infrastructure).
As Phoenix develops COPLINK for their use, the COPLINK prototype must be refined to accommodate additional data such as case narratives, and differences in how the agencies make use of that data. This phase will include database evaluation, surveys of partner agencies, and evaluation of NIBRS standards. In addition, the group will evaluate data storage and retrieval techniques for biometric indicators for possible inclusion into the system design.
The COPLINK team will then prioritize these changes and additions and make design changes based on this information.
As changes are implemented to all levels of the system, the COPLINK team will closely evaluate the initial prototype design and available technologies that will provide the greatest platform and application independence. Refinements to the interface and web application server levels that provide greater ease of development for varied ODBC-compliant databases will be developed and tested on the prototype application.
Deployment of the distributed system will not only gain valuable access to information between the two agencies, but will also provide researchers with a greatly expanded research testbed. The project will address key knowledge retrieval and management technical areas believed to be most critical to law enforcement applications and their organizational impacts: linguistic and neural network based case analysis, information visualization; and law enforcement search agents.
In Year 2, we plan to experiment by applying the Arizona Noun Phraser to law enforcement case records. The team will then apply developments in linguistic analysis to summarization and visualization techniques, and to the law enforcement Concept Space. Concurrently, the team will begin development of a prototype law enforcement search agent/spider for Internet and COPLINK Intranet information sharing.
Our earlier COPLINK experimentation shows that significant improvement will be required to make the phrasing techniques useful in law enforcement applications. In particular, selected entity extraction techniques need to be added to the Arizona Noun Phraser in order to extract semantics-rich objects, e.g., vehicle make/model, people’s names, place names, organization names, etc. Such entities will convey a much richer context for further analysis. Instead of simply finding relevant noun phrases in the collection, we plan to develop several entity extractors by incorporating several domain-specific entity lexicons, e.g., the US Geological Survey’s (USGS) Geographical Names Information Systems (GNIS, a large gazetteer containing 1.8 million place names), the EDGAR company name lexicon (electronic filing of US companies), etc. Such expansion will allow us to better represent different types of objects in the law enforcement applications.
Techniques such as noun phrasing can potentially help law enforcement investigative personnel extract information from very large collections of report narratives and statements, that otherwise is only available through prohibitively time-consuming hand searching and reading by investigators and crime analysts. As these techniques are developed for law enforcement in the COPLINK setting, they can become available to other knowledge management and information sharing applications at every level of the criminal justice system.
Concept space, or automatic thesaurus, is a statistics-based, algorithmic technique used to identify relationships between objects (terms or concepts) of interest (Lesk, 1997). The technique is frequently used to develop domain-specific knowledge structures for digital library applications.
In the University of Arizona Artificial Intelligence Group, the idea of concept space was generated to facilitate semantic retrieval of information. In several user studies concept space was shown to improve searching and browsing in the engineering and biomedicine domains. In the biosciences, one concept space approach was applied to the Worm Community System (WCS) (Chen et al. 1997a) (Chen et al. 1995) and the FlyBase system (Chen et al. 1997b). There also have been successful results in the DLI studies conducted on the INSPEC collection for computer science and engineering (Chen et al. 1998b, Chen et al. 1997a) and on Internet searching (Chen et al., 1998a). Current on-going concept space research is being conducted in geographical information systems, law enforcement, and medicine.
In our ongoing COPLINK project, we have successfully adopted our techniques to create a COPLINK Concept Space based on a collection of 1.5 million case reports from the current Tucson Police Department Records Management System (RMS). These cases span a time frame from 1990 to the present (the entire case record collection for the City of Tucson). Based on careful user requirement analysis, six entity fields from the database were deemed relevant for concept space analysis: person, address, organization, vehicle, weapon, and crime type. The purpose of this tool is to discover relationships between the different crime-related entities.
Criminals are creatures of habit and being able to understand their habits and close associations is an important issue (Joyce, 1997). The COPLINK Concept Space takes advantage of this characteristic by capturing connections between people, places, events, and vehicles, based on past crimes. Our initial evaluation involving 15 actual police searches has shown a dramatic improvement in quantitative performances (time saved from 48 minutes per investigative case to about 18 minutes) and qualitative performances (solving unsolved cases) of the COPLINK Concept Space (Hauck & Chen, 1999). We believe Concept Space will be a scalable and powerful tool for other federal and local law enforcement agencies as well.
Significant research needs to be conducted to investigate using Concept Space with our proposed noun phrasing and entity extraction techniques. In the above example entity fields were identified manually by human analysts. Many law enforcement agencies have begun to incorporate content-rich narratives in their record management systems (e.g., Phoenix Police Department has complete narratives about each case). These narratives will provide a fertile testbed for combining automatic indexing and concept space analysis for intelligence identification.
The concept of using agents is an outgrowth of the past 40 years of research on artificial intelligence and robotics. It has become particularly relevant in the context of Internet research. The idea of a software entity that could perform tasks on behalf of a user was well established by the mid 1970s. (Caglayan, 1997). The research aims to create software to support reasoning, knowledge representation, and learning. The practical applications of agents have attracted significant attention in the 1990s. Many software agents have emerged to support Internet searches, data mining, and collaborative computing applications.
Agents are usually continuously running processes that know what to do and when to do it (Haverkamp, 1998). Csurgay and Heilmann claim that there are six characteristics of agents (Csurgay, 1996) (Heilmann, 1995): autonomy, communication ability, capacity for cooperation, capacity for reasoning, adaptive behavior, and trustworthiness. A number of researchers have explored the use of agents for information filtering, cataloging and delegation. For information filtering, a system needs to proactively survey a large information landscape based on user-specified or system-generated criteria and proactively filter out relevant and timely information, often without a user’s initiation. The system also needs to “learn” from its users, based on prior user preferences and selection pattern (Woolridge, 1998) (Genesereth, 1997). Several new agent-based systems have begun to emerge for the Internet, e-commerce, and database applications.
Agent implementation for law enforcement searches is in its infancy. However, we believe that search agent development for law enforcement promises to be an extremely rewarding research area. Internet crime has become a specialized area for law enforcement investigators, and intelligent agents may greatly facilitate their work. The development of intelligent agents in the COPLINK Intranet context will become increasingly vital as inter-agency information retrieval expands. The opportunity to develop these tools early in the expansion process will allow a timely convergence as both the technology and the span of access mature.
In Year 3, the COPLINK team will conduct lab and user testing of COPLINK linguistic, neural networks and agent/spider technologies. Included in this phase will be the development of an initial prototype in each of the focus areas as well as employing feedback from controlled lab experiments to refine the prototypes.
Based on these evaluations, the group will refine the prototypes and integrate the prototypes with COPLINK databases and Concept Spaces. An important part of this integration is deciding how to integrate the two systems and to what extent. We will work to design the integrated Knowledge Management System based on user requirements as well as task analysis.
During this phase the COPLINK team will conduct field evaluations of the new integrated COPLINK system. Serving as a final testing ground before full deployment, this phase will allow us to deploy the system in a real-life context, while utilizing feedback for refinement. This phase also plays a part in identifying hardware requirements and training issues, which are relevant for the system’s final deployment at Tucson Police Department as well as other law enforcement agencies.
Project Duration: June 1, 2000-June 1, 2003.
Project Deliverables and Timeline (by quarters):
(See Appendix F)
·
Year
1:
(1)
Q1-Q2:
Arizona (Tucson and Phoenix) COPLINK law enforcement databases and concept
spaces creation. All database, web
application severs, and Java front-end will be developed and tested.
(2)
Q3:
Development of a secured, integrated, distributed COPLINK network for Arizona
law enforcement information sharing.
Distributed database design and network protocols will be developed and
tested.
(3)
Q4:
Field studies and assessment of law enforcement information sharing needs and
solutions via COPLINK database and concept space technologies. TPD and PPD personnel will be involved in
actual field studies of the COPLINK system’s effectiveness.
·
Year
2:
(1)
Q1-Q2:
Linguistics-based case (narrative) indexing and analysis. New linguistic parsing rules will be
developed to analyze law enforcement-specific case report content.
(2)
Q2-Q3:
Dynamic, graphical case summary. AI
Group’s acclaimed SOM neural network techniques will be modified to visualize
case report content.
(3)
Q3-Q4:
Development of law enforcement specific Internet agents/spiders for COPLINK
Intranet information sharing (via information pushing) and Internet law
enforcement information filtering. Our
existing Itsy Bitsy Spider will be modified for law enforcement purposes.
·
Year
3:
(1)
Q1:
Lab testing and evaluation of COPLINK linguistic, neural networks, and
agent/spider technologies. Police
subjects will be used in lab experiment to verify the quality of
system-assisted textual mining.
(2)
Q2-3:
Prototype integration of linguistic, neural network, and agent/spider
technologies to COPLINK databases and concept spaces and field evaluation of
new integrated COPLINK knowledge management system.
(3) Q4: Project dissemination. Significant efforts will be placed to disseminate project lessons and systems to other law enforcement agencies.
This project will be evaluated administratively to determine compliance with the stated objectives. This will be done by regularly scheduled status meetings between the Tucson Police Department, Phoenix Police Department, and the UA/MIS Artificial Intelligence group. Both police agency teams will be responsible to keep their respective city IT managers and City Manager’s offices informed of project status. The project team will evaluate the project process and project results once the project is concluded. Results will be compared to the expected or anticipated outcomes and the practical applicability of the research and development project final results.
· The TPD and PPD COPLINK Intranet will provide a showcase for a cost-effective, scalable solution for law enforcement information sharing.
· This showcase will result in expanded participation and implementation of COPLINK technologies by other law enforcement agencies.
· The results of the research will improve and enhance knowledge management and intelligent information analysis and sharing for law enforcement.
· Increased efficiency and improvement of law enforcement access and use of investigative information will improve law enforcers’ effectiveness within their communities. This will ultimately benefit the communities and the public that the agencies serve.
The fully deployed and operational distributed COPLINK system will provide investigative and field personnel with more effective tools and information for the apprehension of criminals across jurisdictional boundaries.
Future COPLINK Information
Integration Plan: Federal and Court Systems
The COPLINK project must continue to broaden the extent of implementation and to integrate other data sources if possible. The first priority is to refine the application in a narrower (state-wide) scope to implement a proof-of-concept system. Future utilization of COPLINK technologies, given adequate time, funding, personnel, expertise, and equipment will potentially broaden into the following areas: technology, functionality, and number and type of databases.
Any interoperability project
for law enforcement must consider the criminal justice system as a whole. Information sharing at the national level
exists now through such national level databases and networks as NCIC (National
Crime Information Center) and NLETS (National Law Enforcement Telecommunications System), as well
as state criminal history databases such as ACIC (Arizona Crime Information
Center) and Interstate Identification Index
(III). However, information
stored at a national level must be highly summarized and selective. Extremely valuable case and intelligence
detail information will likely remain at a local or regional level. Information sharing at the agency level must
complement the resources and systems in place and under development at the
federal and state level. COPLINK
research will complement current central federal database development efforts
by supporting grass-root information and knowledge management needs of local
and state law enforcement agencies. As
COPLINK technology expands locally, the COPLINK technology team will identify
innovative ways to expand vertically as well, to collaborate with state and
federal agencies in providing comprehensive knowledge management solutions at
all levels.
Horizontal
expansion of agencies and data sources must consider court systems at all
levels of government, as well as state motor vehicle registration bureaus. COPLINK supporters are currently working
with state level information technology agencies such as ACJC (Arizona Criminal
Justice Commission) and GITA (Government Information Technology Association) to
facilitate statewide information technology planning efforts (See Appendix
E). With its continued support from the
National Institute of Justice, COPLINK will work at the federal level as well
to solidify its role in the national law enforcement infrastructure.
Upon
completion of the research and an action planning process, a final report will
be prepared discussing the project and its outcomes. The Tucson Police Department will mail copies of the final
research report to large law enforcement and criminal justice agencies
nationwide. This mailing will include
information on how TPD and PPD intend to apply the technology research to
prevent crime, apprehend criminals and to enhance public safety. UA/MIS Artificial Intelligence Group will
also make use of the research and development findings in other applications
and may market the technological processes developed as a result of this
award. Additionally, the Project Managers
will author one or more articles for submission in technical and professional
law enforcement journals discussing the development, findings, and practical
application of this research. Copies of
all disseminated information will be forwarded to the National Institute of
Justice.
Authored by: Sgt. Jennifer Schroeder, City of
Tucson Police Department
Dr.
Hsinchun Chen, UA/MIS Artificial Intelligence Group
Roslin
V. Hauck, UA/MIS Artificial Intelligence Group