Nanos gigantium humeris insidentes!

KiwiPredator Architecture

  • July 16, 2013 12:05 am

Feel free to check out my repo at





Parser is the component that converts sentence into a parser tree. Which is done by Stanford Parser. 


Extractor is the component that extract subject, predict and object from a sentence. It contains two main parts — a markup language called tree-based regular expression language (TREL), its interpreter and a tregex engine which perform regular expression on a parser tree.

TREL has two characteristics.

  • Markup Language
  • Inheritance

There are three main types of ERML. Splitting, Pruning and Replacing.

  • Splitting: Mainly used for splitting one tree into several subtrees.
  • Pruning: Mainly used for pruning uninterested parts.
  • Replacing: Mainly used for replacing pronoun, tense and number type.


Evaluator reviews the hypothesis and either takes it as a new item of knowledge or discard it. In either case, it will send feedback to the TREL. 

Print Friendly