A Probabilistic Approach for Integrating Heterogeneous Knowledge Sources

Published on 2014-07-301961 Views

Arnab Dutta

Open Information Extraction (OIE) systems like Nell and ReVerb have achieved impressive results by harvesting massive amounts of machine-readable knowledge with minimal supervision. However, the know

ESWC 2014 - Crete

Related categories

Presentation

A Probabilistic Approach for Integrating Heterogeneous Knowledge Sources00:00

Which Superman exactly?00:09

Uncertainty in the Web00:57

Information Extraction (IE) Projects - 101:21

Information Extraction (IE) Projects - 202:06

Approach02:57

Approach: Baseline03:15

Approach: Baseline (contd.)04:05

Baseline drawbacks04:33

Approach: Probabilistic04:49

Big Picture05:18

Methodology06:10

I. Probabilistic Type Generation06:58

Tree Generation07:43

Node Scoring09:38

Tree Generation (α-tree)10:04

II. Formulate as a Markov Logic Network11:49

Grounding the Markov Network14:26

Can we still do better?16:06

Bootstrapping16:13

Bootstrapping in action16:21

Experiments18:01

Learning α18:14

Variation of scores with α 18:42

Comparative Values (α = 0.5)18:49

Limitations19:24

Looking ahead19:49

Thank You20:10