MetaPAD

Meta Pattern-driven Attribute Discovery from Massive Text Corpora

Datasets

NYT News Dataset

zip/data-metapad.zip typed news corpus

You can check your output with zip/output-metapad.zip meta patterns after segmentation (top or bottom) matching tables with meta patterns of appropriate granularity (top-down or bottom-up) results of information extraction using meta patterns (attribute)

Newsbank Dataset

Entity linked done by matching to Freebase (thus possible typing errors).

Data is in the input/ folder.

Requirements

This version should be compatible with Python3. C++ code requires OpenMP library to run.

How to run

Execute run.sh

Code

Encryption: metapad.py
pattern segmentation and classification : C++
synonymous pattern grouping: Python metapad.py
type level adjustment: Python metapad.py