This repository is an enhanced version from the forked library. It is an implementation from the publication PATTY: A Taxonomy of Relational Patterns with Semantic Types. This implementation can have slight modification from the original paper due to limited information disclosed and the results are not guaranteed.
The goal of PATTY is given a entity labed corpus, extract entity relationships and locate patterns of sentneces for future entity extraction usage.
PATTY requires python 3.5+ to run. All external libraries are specified within the requirements.txt file.
python3 -m venv .venv
pip install -r requirements.txt
PATTY requires two user input information: entity and corpus data.
- entity: located in
src/entity.py, it specifies the entities that were to be extracted and appears in the corpus - data:
data/requires a file that follows the following structure. Each line is a document that entities are highlighted by appending<Entity>_in front.
PERSON_Tom_Thabane resigned in DATE_October_last_year to form the All Basotho Convention (ORGANIZATION_ABC), crossing the floor with NUMBER_17 members of parliament, causing constitutional monarch King PERSON_Letsie_III to dissolve parliament and call the snap election.
To execute the program, run:
cd src
python main $DATA_ROOT $CORPUS_NAME
All output files are in binary for the intrest of data size and machine readability. It can be transformed into text by running transform.sh in the data/ folder.
./transform.sh