System Architecture

A thorough overview of just the algorithms and system architecture is done in the fourth chapter of the thesis ( http://reports-archive.adm.cs.cmu.edu/anon/2014/CMU-CS-14-128.pdf).

Briefly, the following text diagrams show each of the major components:

Text Processing

Text  ==>  Document  ==>  Candidate  ==>  Feature Vector

Training Data Creation

                                        Knowledge Base Facts
                                                             \\
Query List  ==>  Indexed Documents  ==>  { Text Processing }  ==>  Labelled Data

Relation Extraction Learning

Labelled Data ==>  Cost-Sensitive Binary SVM Learning  ==>  Parameter Sweep + Selection of Best on Held-out Data  ==>  REx Model