QA3tagger - A tagger (entity recognition) tool for the Question Answering over Data Cubes (QALD6 Task3)

Introduction

This repository contains the tagger used by QA3 (QAcube), awarded 2nd place at the QALD6 task3 challenge.

Setup

From main dir, run the following to download dataset dumps:

wget http://linkedspending.aksw.org/extensions/page/page/export/qbench2datasets.zip
unzip qbench2datasets.zip 
#optionally: rm qbench2datasets.zip

Also run:

pip install -r requirements.txt
python -m nltk.downloader stopwords punkt wordnet

Run

Run python webserver.py and go to [http://127.0.0.1:8080/qa3?q=How much was spent in Nigeria?](http://127.0.0.1:8080/qa3?q=How much was spent in Nigeria?). It depends on python Flesk.

You can also use Docker:

# may take more than minutes (!) and produces a large (5Gb) image
# this is mainly due to download of qbench2datasets and index creation
time docker build -t atzori/qa3tagger .
#or: time docker build -t atzori/qa3tagger https://bitbucket.org/atzori/qa3tagger.git


# run container
docker run -p 8080:8080 --name qa3tagger atzori/qa3tagger