Relational Learning With Expert Advice

Document Type


Publication Date


Published In

Text Analysis Conference Workshop


The introduction of the paper, in lieu of an abstract: We present the Wisconsin-Indiana system for the TAC KBP 2015 slot filling task. The proposed approach is a three-staged pipeline. The first stage consists of converting the raw text to a large set of rich relational features (we employ first-order logic for representing the underlying text). During the second stage, weak supervised examples are generated using expert human advice that form weak annotations for the text. In the final stage, the relational features and weak labels are provided as input to an efficient learning system [3] that learns a Relational Dependency Network [6] from the given data.


Text Analysis Conference Workshop (TAC 2015)

Conference Dates

November 16-17, 2015

Conference Location

Gaithersburg, MD