In The Past

Most computerized real estate databases or indexed file systems treat legal descriptions, names and addresses essentially as unusable text .They could print the descriptions that were in the database, but they could not find a piece of property with a particular description .This is because traditional computer databases treat text information as just a series of letters and numbers; they do not try to deal with the meaning of the text.In the ensuing discussion we shall focus primarily on the RDI LDAMS with examples concerning the RDI name and address matchers occurring in natural progression within the main discussion.

Hence, in order for the computer to find a record with a certain legal description, the program had to know, in advance, the exact order of words in the description, and possibly the exact spelling too.For example, if your database contained a record with " LT 23 BLK 3 MARINA BAY PARK SEC 1 " , and you were looking for " BLOCK 3 LOT 23 MARINA PARK SECTION 1", the computer would not find the record, because the words are in the wrong order, and the word "BAY" is missing from the example.Even if you got the words in the right order, it still would not find the record, because "LT" and "LOT" have different letters.

In the past, it was difficult to make computers deal with the meaning of legal descriptions because of the many different ways descriptions can be constructed. Words can be abbreviated or misspelled; word order can be changed; ranges of numbers (such as "5-8" ) can be specifically  enumerated (such as "5,6,7,8. ); one description may have survey, subdivision, and abstract while another has only the subdivision name.Furthermore, many descriptions are not the nice orderly "lot/block/section" type of descriptions.

Because of these limitations, in order to find the records for a piece of property in a traditional database, you had to have some unique identifying information, such as the tax ID number assigned by the county appraisal district, or the address.Many documents dealing with real estate, given the multiple sources of data, do not supply that information, so what can you do then to match non-unique sources to one another based on mere free text in absence of meaning?

Even if you have an address, it is frequently difficult to find a matching property record, because of differing abbreviations (STREET vs. ST), or misspellings (WESTHIMER vs. WESTHEIMER), or alternate street names (NORTH FWY vs. IH-45 N).As with legal descriptions, traditional search programs may treat street names as just a series of letters, and would have difficulty dealing with these examples.

Legal Descriptions Are More Than Just Text

Using RDI. s Legal Description and Address Matching System (LDAMS), it is possible to use one legal description from one database, such as a Assessor. s Database, to locate another legal description in another database, such as a County Clerk (Recorder) Database, that means the same thing, even if the legal descriptions are typed differently, contain different abbreviations, punctuation, etc.

For example, the LDAMS would know that the two examples given earlier actually mean the same thing.For many legal descriptions, the LDAMS can find a single record in the database with the specified description; in other cases, it may narrow the list of possibilities to a few records, which then applies additional rules in its Expert System Rules Database to determine which one is correct.

Obviously, the matching is from one database into/onto another one, although it is also possible to operate the LDAMS in a multi-threaded mode.

How It Works

The RDI Legal Description and Address Matching System uses a technique derived from Artificial Intelligence (AI) called "Rule-Based Programming" to emulate the knowledge of a person expertly familiar with Real Property Legal Descriptions. Using this technique, a set of  "rules" defines how meaning can be deduced from the text of a description. For example, the program has rules that indicate that "TRACT 3-5" is a list of three tracts while "TRACT 3-A" is a single tract.It has rules that indicate that lots and tracts typically exist within blocks (but not always).It has rules that indicate that the phrase "MARINA BAY PARK SEC 1" refers to a subdivision name Marina Bay Park, even though the word "SUBDIVISION" does not appear in the phrase.These rules are interpreted by an artificial intelligence tool, which was developed at NASA/Johnson Space Center in Houston (CLIPS).The rules allow the AI tool to derive the meaning from a legal description, and to determine when two different descriptions can mean the same thin

Figure 1 shows how RDI has used the LDAMS to automate part of the process of connecting Harris County Clerk Deed records to the corresponding tax records from the Harris County Appraisal District.

The LDAMS begins by converting all of the descriptions in the Harris County tax records to a standard form based on their meaning, and saving those standard forms for later use.Then, when a new deed record is available and we want to find it's corresponding tax record, the LDAMS uses the legal description and grantor/grantee names from the deed records to locate candidate matches from the tax database.The candidates are then compared to the deed record and each other to isolate the best possible match. There are no common unique indices between these two sources, like, say, an APN or Tax IDs. (Even with a unique APN or Tax ID, data entry errors reduce this uniqueness by at least 3-5% so the LDAMS fits in perfectly in these cases to advance the overall match rate).