Thales provides a Java full-text resource indexer and search engine. Multiple search contexts may be created. Any LDBC-supported database may be used including MySQL, Oracle, MS-SQL, HSQLDB and DB/2.

Thales differs from other "bag-of-words" indexes in that there may be multiple bags (contexts) and the search may be limited to one or more of them. The Thales engine indexes text and associates the text with a reference. Each reference has a title. Thus an engine to index and search an entire web site might to read each page on the site, take the title element and the URL to create a reference and reference title and then index the textual contents of the page. In addition if the author was specified the authors name might be indexed in the "author" context.

Thales uses JDBC and the LDBC package. The JDBC driver must be one of those supported by LDBC.

Thales Concepts

Words and phrases are indexed within the scope of a Reference. They may further be restricted to a single Context within the reference.

Reference

An object being indexed. Each reference has a name and a title. The name is generally a machine-readable reference while the title is the human-readable title for the same item. Thales does not store the actual referenced item just the name and title. So Thales could be used as a catalog of images where the reference name is the actual image URL and the title is a brief description of the image. The text to be indexed would be a long description and perhaps some information about the creator of the image.

Word

Words and/or phrases indexed as whole items. Thales allows the indexing of phrases as complete items. It is up to the application to be able to recreate the phrases for the query.

Content

The context in which the word is being used. For example the description of an image or the author of a document.

Querying

When querying the database the context(s) that are to be searched may be specified or, if null, all contexts will be searched. For example you may want to index documents and index the body of the document in the "body" context and the authors in the "author" context. This would allow a query to locate all documents that John Smith had authored separate from all the documents about John Smith the author.

Words are located by direct hit, synonym hits and soundex hits. The relative values associated with each hit type are set in the ThalesConfig object. Setting a value to 0 (zero) removes it from the calculation. The relative values of each word along with the number of words in the query and the number of times the word appears in the database factor into a final rank for a resource. With the highest rank being the closest match to what was requested. The results are sorted by rank order.

Finding

When finding words only exact words are located. This is much quicker than the Query operation but is much more restricted.

Cross Referencing

Once a reference has been located it is possible to find all other references that share the same indexed words. The XRef mechanism provides this functionality.

Thales Requirements

• A Java runtime (JRE) version 1.3.1 or above. A complete Java SDK is needed to compile the source code.
• A JDBC driver and the LDBC package is required. The JDBC driver must be one of those supported by LDBC.
• Jakarta Log4J for flexible log file handling.
• The Xenei Utilities.

Optional Packages

• Jakarta Ant version 1.5.2 or above if you plan to compile the source code.