Skip to content

GraphParser is a semantic parser which can convert natural language sentences to logical forms and graphs.

License

Notifications You must be signed in to change notification settings

sivareddyg/graph-parser

Repository files navigation

GraphParser

GraphParser is decribed in the paper http://sivareddy.in/papers/reddy2014semanticparsing.pdf

Installation

GraphParser is written in Java, and requires few external java libraries. You can install them using

./install.py ungrounded

Ungrounded Semantic Parser

GraphParser can parse natural language sentences to logical parses and graphs. Run the following command

cat input.txt | sh run.sh

Online demo

Online demo of GraphParser can be accessed at http://sivareddy.in/graphparser.html

Evaluation datasets

If you are interested in evaluation datasets, and the output of GraphParser on the test datasets, you can download them using

./install.py evaluation

The datasets will be downloaded to the folders data/tacl_splits and data/tacl_ouput.

Grounded Semantic Parser

To replicate TACL results, you will have to install Freebase SPARQL endpoint. Please email me personally.

References:

If you are using GraphParser, please cite

@article{reddy_largescale_2014,
  author = {Reddy, Siva and Lapata, Mirella and Steedman, Mark},
  title = {Large-scale Semantic Parsing without Question-Answer Pairs},
  journal = {Transactions of the Association for Computational Linguistics},
  volume = {2},
  year = {2014},
  pages = {377--392},
  url = {http://aclweb.org/anthology/Q14-1030}
}