SPRL-Spacy

This repository implements an easy to use Spatial Role Labeling module trained on three entities (TRAJECTOR, SPATIAL_INDICATOR, LANDMARK) and the relations appearing on the SpRL 2013 IAPR TC-12 dataset.

Requirements

• spacy >=2.0.0a18 and the necessary requirements (Does NOT work with Spacy versions 2.1.x and above)
• sklearn
• scipy
• pickle for python 3.7.0
• problog for use with ProbLog.

Usage

1. Clone this repository where you want to use it.
2. Download the two models from the releases page and put them in the models/ directory.
3. Import spacy and sprl and use them like the following example:

import spacy
from sprl import *

nlp = spacy.load('models/en_core_web_lg-sprl')

sentence = "An angry big dog is behind us."

rel = sprl(sentence, nlp, model_relext_filename='models/model_svm_relations.pkl')

print(rel)

If everything went fine you should get something like:

[(An angry big dog, behind, us, 'direction')]

You can also run sprl_cmd.py to get a continuous input to test how well various sentences are processed:

$ python3 sprl_cmd.py

Problog

If you happen to have problog installed, I have made a library that allows you to process sentences and produce a set of first order predicates that express the spatial relations within it. For example you can do something like in pl/test_sprl.pl:

:-use_module('sprl.pl').

run_all :- sprl_process_sentence('An angry big dog is behind us.').
query(run_all).
query(trajector(X)).
query(landmark(X)).
query(spatial_indicator(X)).
query(type(X,Y)).
query(extent(X, Extent)).
query(spatial_relation(X)).
query(gtype(X,Y)).
query(srtype(X, Y)).
query(srtype(X)).

which you can run with:

$ PYTHONPATH="../sprl" problog test_sprl.pl

and get the following output:

              extent(lm0,us):   1
          extent(sp0,behind):   1
extent(tr0,An angry big dog):   1
        gtype(st0,direction):   1
               landmark(lm0):   1
                     run_all:   1
      spatial_indicator(sp0):   1
       spatial_relation(sr0):   1
             srtype(sr0,st0):   1
                 srtype(st0):   1
              trajector(tr0):   1
            type(lm0,person):   1
            type(tr0,animal):   1

We can see for example that it identified and labeled the trajector, landmark and spatial indicator in the sentence, assigned them an id, identified the spatial relation and assigned it a general type of direction. It also assigned a type of person to the landmark us and animal to the trajector An angry big dog. For what those predicates mean and how they are used please see doc/sprl.html.

Credits

While the model has been trained by me, the relation extraction part uses features from the paper for Sprl-CWW (see below), and the dataset from SemEval 2013 Task 3: Spatial Role Labeling.

The features for relation extraction:

Nichols, Eric, and Fadi Botros. 
"SpRL-CWW: Spatial relation classification with independent multi-class models." 
Proceedings of the 9th International Workshop on Semantic Evaluation.

Semeval 2013 task 3: Spatial Role Labeling

Kolomiyets, Oleksandr, et al. 
"Semeval-2013 task 3: Spatial role labeling." 
Second Joint Conference on Lexical and Computational Semantics

So please cite the papers above, as well as spacy and ProbLog (if you use it) in your work :)