ASLN 2016: Question Answering with a AMR knowledge base.
Ask question about the Little Prince and get the answer. Implemented using AMR and the Little Prince corpus from the Linguistic Data Consortium.
Tested with Python 2.7.12 (CAMR still not ported to Python 3) and Ubuntu.
Before prociding make sure CAMR: A transition-based AMR Parser is working on your computer (https://github.com/c-amr/camr). A requirements.txt file is provided to help you install CAMR dependencies, but is not necesarry if is already installed. Apart from CAMR, this project doesn't have any dependency.
Important: Clone CAMR Github repository in the root folder (AMR-QA/camr). Then place the model file (as describe in CAMR repository) inside the carm folder with the name 'LDC2014T12.m' (AMR-QA/camr/LDC2014T12.m)
Finally, run it:
$ python QA.py
Notice: python 2 path should be accessible through python, not python2 command. If not use Virtualenv or Pyenv and make sure that when calling:
$ python --version
you get the output:
Python 2.7.***
You can try the phrases: Who said? Who run What said
Questions should include one and only ones of the following question phrases and searchs for one of the second arguments:
- for who -> ARG2
- how long -> duration
- from where -> origin
- to where -> destination
- who -> ARG0
- what -> topic, ARG1
- when -> time
- where -> location
- how -> manner, instrument, mode
- why -> purpose