As the title suggests, this paper describes techniques to exploit parallelism of the Connection Machine for free-text search algorithms while searching natural-language documents contained in large databases, ranging from a few megabytes to a few gigabytes. It also provides an overview of the Connection Machine architecture. The authors then describe an algorithm for free-text searching based on “surrogate coding” techniques. The paper has excellent illustrations describing “surrogate coding” techniques. The drawback of “false-hits” and a technique to minimize them are described.

Then, the paper describes techniques to implement the algorithm in the Connection Machine in order to exploit its parallelism. Even though the authors have repeated the concept of “Boolean Queries” and “Simple Queries” in two different places, I found the explanation useful. The paper has also provided a very good description of the “relevance feedback” concept in order to formulate effective “Simple Queries.” Finally, the authors describe the implementation of these concepts on a Connection Machine and provide some benchmark information on attributes, such as “precision,” “recall,” and “response time.” The paper was very well written, well organized, and, thus, easy to read. This paper is must reading for those interested in efficient text search algorithms for large databases.