Changing the Rules of the Game

Solving a crime can be like playing the board game Clue: The key is to take lists of names, weapons and places and find the links among them, creating new leads or a short list of suspects to track down.

Unfortunately, as with Clue, information valuable to a case is often in different hands, either stored in separate databases within a department or in different jurisdictions altogether. An automobile or weapon involved in one case, for example, might be associated with a suspect in a past case, yet the police may not find the link.

But city police departments in Tucson and Phoenix, Ariz., are trying to change the nature of the game with a project called CopLink.

The initiative, funded in part by a grant from the Justice Department's National Institute of Justice, aims to make it easier to uncover those hidden links by creating an Internet-based system for integrating and accessing that data — essentially, putting all the cards on the table.

It also sets the stage for an even more intriguing development, in which the police can improve their ability to analyze that data by borrowing some techniques from the government intelligence community.

Managing Knowledge

CopLink, under development for more than two years, builds on the information systems that many police departments have installed over the last several years to manage their case reports, as well as their various databases of names, weapons, automobiles and similar information.

The police keep such information indexed for future use, because people, places and objects involved in past cases tend to appear again in other incidents.

Tucson police, for example, developed "better than average" records management systems, said Sgt. Jennifer Schroeder, the city's CopLink project manager. But, as in other cities, these efforts typically produced "isolated systems in the department and a lot of different data sources that officers had to look at to find information."

In 1996, when the Tucson Police Department began looking for a way to fix this problem, a colleague studying for his Management Information Systems degree introduced Schroeder to one of his professors, Hsinchun Chen, director of the artificial intelligence laboratory at the University of Arizona in Tucson.

Chen had tackled similar "knowledge management" problems encountered at numerous government agencies, including the CIA and the Defense Department's Advanced Research Project Agency.

Knowledge management generally refers to methods used to track and analyze all the information stored in an organization's various data sources. The key is to extract the information from those sources, consolidate it and then index it so it is fully searchable, Chen said.

Once that work is done, an organization can begin "squeezing out the intelligence," Chen said.

Working together on a proposal for an NIJ grant, the Tucson police and the university developed a similar strategy for integrating police records. NIJ funded the project in 1997, and the Phoenix police department signed on after seeing a demonstration of a CopLink prototype.

CopLink is designed to bring together different types of information that might be used to investigate a case.

For example, the police may have witnesses who saw a red Corvette, driven by a male between 18 and 25 years old, leave a crime scene. The investigator can cross-reference data on both people and automobiles to narrow the search, then download mug shots of those people to show to witnesses.

Without the integration, that can be time-consuming, said Joe Hindman, computer services administrator for the Phoenix Police Department.

The idea is not to integrate the original applications — they run without interruption. Instead, CopLink integrates the data, extracting it from the applications and creating a new database or "data mart." This serves the same purpose but eliminate the technical complexity of trying to combine applications that were developed separately.

Tucson and Phoenix plan to have their basic systems in place by the end of the year. Both departments plan to start with just a couple of databases, adding others further down the road.

Police will be able to access CopLink through a secure Internet link using a simple point-and-click Web browser interface. Because the system is Internet-based, law enforcement personnel eventually could access the data from any place that has an Internet link — including a squad car equipped with a laptop computer and wireless modem.

Likewise, the Internet provides an excellent tool for sharing information across jurisdictions.

Tucson and Phoenix, for example, plan to establish a secure Internet link between their CopLink systems. The two cities account for about 65 percent to 70 percent of the cases in the state and now will be able to share their data, Hindman said.

They said they hope other departments in major population areas will set up their own CopLink systems and extend the network, through either intranet or extranet connections. Additionally, departments in large cities could also work with smaller departments in their own region to combine data, as Phoenix is doing (see related story).

In such a scenario, each system essentially serves as a CopLink node, maintained independently but accessible to all CopLink users. Police will be able to compare information pulled from their own system with data stored in a neighboring CopLink node.

The overall CopLink network will support any number of nodes without suffering performance problems because each node has its own processing power. "We have looked at quite a few applications, and this is the most complete and scalable solution for law enforcement," Chen said.

The Payoff

Although the initial CopLink systems will not be in place until late this year, both Tucson and Phoenix are looking forward to the next phase of the program, when the real payoff could come.

The key to CopLink, of course, goes back to Clue: Police solve crimes by finding links among different types of data, such as names, weapons and places. Theoretically, the more data available, the more links might be found. But the complexity of searches also increases with higher amounts of data. So the next logical step is to automate database searches.

Chen, in his work for the federal government, has done just that. Chen specializes in a field of AI known as text mining. Text mining software "reads" and analyzes documents, one word at a time, using linguistic analysis to identify revealing patterns in the text.

The software is able to read through a large amount of information very quickly to turn up links among different text elements that a person, working alone, might not have found, such as a particular group of people using particular weapons in a particular place.

In one of his early projects, Chen's program analyzed information in the CIA's Russian computing database, looking for links that revealed which particular scientist developed a particular project, or what offshoots that project produced. Later, he worked with the National Cancer Institute to apply the software to cancer research, reading through about one million medical abstracts to identify key developments in medical protocol, gene research and other topics.

Unlike standard databases, which provide only exact answers, this software "suggests" possible links that can be tested with additional searches. In successive searches, a person can use a process of "triangulation" to uncover increasingly specific information, Chen said.

For example, an investigator may have information on a few names or a weapon. He can enter that data into the system and see what associations appear in the database. Those associations serve as leads for further research.

The software will be especially useful in working with the "narratives" filed along with standard reports, Hindman said. "Officers do bury some important information in the text, and this information is very valuable in developing case leads," he said. Going to Market

The people working on CopLink see the potential for this system to take the law enforcement community by storm. They are not alone.

A group led by Chen recently founded Knowledge Computing Corp. to market this technology to other law enforcement agencies and to the private sector. Earlier this year, the company received $1.5 million in venture capital funding.

CopLink "has the potential to be the national model for law enforcement information sharing," said Chen, who is especially interested in seeing applications where the police access CopLink through cellular networks.

"I am hoping in the next five years that 50 percent of agencies in the nation will be using CopLink," he said.