As homeowners chased low mortgage rates in 2003, county clerks were hit with a deluge of paper. Each time a property changed hands or someone refinanced a home, a county office recorded the transaction and indexed the relevant documents, so they could be easily retrieved if needed.

At the height of the summer 2003 rush, Alameda County, Calif., recorded 2,500 to 4,000 property documents per day. "We found ourselves having difficulty keeping up with the volume, with limited resources," said Malinda Jones-Williams, division chief in the County Clerk-Recorder's Office in Oakland.

Alameda County uses the Anthem document recording system from Hart InterCivic of Austin, Texas, to manage property documents. As part of the recording process, county employees scan documents into the system, where they are stored as electronic images. When someone needs a document, an employee retrieves it by searching a database for a key data element, such as the buyer's or seller's name, or a number identifying the property. Members of the public also can search for indexed public records on the county's Web site.

Although Anthem automated many aspects of property records management, getting those data elements into the index remained a time-consuming manual process. "Someone had to sit there, look at the document and extract certain required information," Jones-Williams said. That meant knowing where on the document to look for key data elements and accurately typing each element into the appropriate field.

Under the county's "key-verify" procedure, one clerk examined the scanned document and typed the index data; a second clerk then viewed the same image and repeated the indexing process. If the two entries matched, the system accepted the record. If they differed, the second clerk had to reconcile the problem and make sure the right data went into the system.

A Struggle to Keep Up

"We were getting close to eight weeks behind," Jones-Williams said. Even after adding three temporary workers, the office struggled to keep up with the load. Because the indexing wasn't up to date, the public didn't have access to the latest and most accurate information, she said.

In November 2003, Alameda County started using an Anthem function designed to expedite indexing. After several months of running Hart InterCivic's Automated Indexing module, the office was no longer behind, Jones-Williams said. "We are verifying documents that were recorded yesterday. We've gone from close to eight weeks to one day."

Automated Indexing uses expert systems technology to pluck the correct data from a scanned document and automatically enter it in the index database. The engine that drives this function is aiINDEX, developed by Mentis Technology Solutions of Centennial, Colo. Hart InterCivic started offering Automated Indexing as part of Anthem in 2002. The first customer to implement the module was Snohomish County, Wash., in early 2003.

The first step in the Automated Indexing process occurs when optical character recognition software translates the printed document's scanned image into machine-readable text. Many systems perform this kind of translation, said Matt Walker, vice president of eGovernment Solutions at Hart InterCivic. The trickier task is analyzing free-form documents like property deeds to find key text strings, and then determining which strings belong in which fields.

Automated Indexing uses metadata to imitate what clerks do when they find the information, Walker said. Gradually the system learns to think like a trained indexing clerk. "Through repetition and through a lot of historical information, feeding in documents and saying, 'Here are the fields I care about on this document,' the system learns and gets better with more statistical volume."

Clues that help the software spot relevant information vary from document to document and from county to county. On one type of deed, for example, the system might find the seller's name just after the word "Grantor" followed by a colon. On another, the tip-off

Merrill Douglas  |  Contributing Writer