- Use python 3.10+
- Install the
requirements.txt
usingpip install -r requirements.txt
- Run file
FOL_solver.py
and wait until the file is done.
1.1. This file runs the CCG parses on all the sick-nl sentences and then solves only for the Trial cases.
1.2. This is currently in windows mode, to change to linux mode open the file and change the line:mode="windows"
tomode="linux"
.
1.3. To change from the trial dataset to the test dataset, change the lineif dataset != "TRIAL":
toif dataset != "TEST":
, same applied for TRAIN. - check the folder
output\
for all the results.
2.1.test_all_read.png
contains all the trial code cases where errors are labeled as neutral.
2.2.test_parse_read.png
contains all the trial code where errors are omitted.
2.3solve.py
see below.
The first line always includes the problem number and the true label from the sick-dataset. The next three lines contain information about the FOL strings.
H is the hypothesis.
P is the premise.
S is the solution.
If either the Hypothesis or Premise is an error, the solution is also automatically an error
.
If the H or P contains a Lambda, the prover9 sread_term
error is provided.
If a FOL string is incorrectly parsed, the prover9 sread_term
error is provided.
If the parsing is succesful S will be in the form of one of three cases;
s:True | False -> Entailment
s:False | True -> Contradiction
s:False | False -> Neutral