I’ve discovered Fluid Construction Grammar which has both a parser and generator implemented in lisp. I am now writing a java implemenation of the FCG core to replace my current construction grammar and parser. FCG accomodates construction composition using feature structures and unification. One of the FCG authors, Joachim De Beule has offered encouragement.
Progress will be tracked on the project’s SourceForge feature tracker.
I’ve sifted through my bookshelves to find the texts that have most influenced the design and implementation of this project to create artificial intelligence. Fifty-one are listed here and and a convenience link is present in the sidebar. Many of these texts are currently more expensive than when I first purchased them, and for the frugal among you, most should be available at your nearby university library.
The RDF Entity Manager uses Java source code annotation to persist objects to the Sesame 2 RDF store. I have packaged the component and it is available for download at SourceForge. The package contains all the required jar files, including Sesame 2 beta 5. Samples demonstrate how to annotate Java classes, persistence and find methods invloving a Sesame repository.
This is the third generation of the semantic annotation component. My first version, the DomainEntityManager, was created in the fall of 2006 and used MySQL for a faithfull implementation of OpenCyc in a relational database schema. Because of a lookup slowdown at 20 million assertions, I reimplemented the relational database using Oracle Berkeley DB Java Edition, which sacrifices SQL compatibility for speed. But the Berkeley DB used much more disk space than MySQL, and that led me back to first principles. By sacrificing full OpenCyc compatibility, and using only the RDF compatible version of OpenCyc content, I was able to consider RDF triple stores for the project knowledge base. Sesame (OpenRDF.org) is a leader and I am very happy with its space utilization and performance. I am able to partitition the KB and cache a partition completely in Linux RAM (tmpfs) for full CPU utilization because of no disk I/O waits. The RDF Entity Manager uses only Sesame right now, but it could be adapted to other triple store implementions, especially if they support Named Graphs (context).
