From 54d3f347875ac27336212d40ae7b500417caf606 Mon Sep 17 00:00:00 2001
From: Joel Oksanen <joel.oksanen17@imperial.ac.uk>
Date: Wed, 25 Mar 2020 20:22:41 +0200
Subject: [PATCH] Noun phrase parsing now returns tree instead of string. TODO:
Use the tree in opinion target replacement
---
.../SemEval-2016/prep_data.py | 73 ++++++++++++++++---
1 file changed, 62 insertions(+), 11 deletions(-)
diff --git a/ADA/aspect-based SA data/SemEval-2016/prep_data.py b/ADA/aspect-based SA data/SemEval-2016/prep_data.py
index a7e93a4..0c8133c 100644
--- a/ADA/aspect-based SA data/SemEval-2016/prep_data.py
+++ b/ADA/aspect-based SA data/SemEval-2016/prep_data.py
@@ -4,7 +4,7 @@ import string
from nltk.tree import Tree
-included_labels = ['NN', 'NNS', 'NNP', 'NNPS', 'DT', 'CD']
+included_labels = ['NN', 'NNS', 'NNP', 'NNPS', 'DT', 'CD', 'FW']
# Marks all subtrees with descriptive label with 'DESC'
def get_np_words(np):
@@ -19,6 +19,22 @@ def get_np_words(np):
return [np_sub]
return w
+def get_np_tree(np):
+ children = []
+ for np_sub in reversed(np):
+ if type(np_sub) is Tree:
+ if np_sub.label() not in included_labels:
+ return (Tree(np.label(), children) if children else None, False)
+ else:
+ subtree, cont = get_np_tree(np_sub)
+ assert subtree != None
+ children = [subtree] + children
+ if not cont:
+ return (Tree(np.label(), children), False)
+ else:
+ children = [np_sub] + children
+ return (Tree(np.label(), children), True)
+
def filtered_np(np):
w = get_np_words(np)
i = (len(w) - w[::-1].index('DESC')) if 'DESC' in w else 0
@@ -31,8 +47,10 @@ def extract_extended_nouns(tree_str):
for tree in trees:
for subtree in tree.subtrees():
if subtree.label() == 'NP':
- # np = ' '.join(np.leaves())
- phrases.append(filtered_np(subtree))
+ np, _ = get_np_tree(subtree)
+ if np:
+ np = ' '.join(np.leaves())
+ phrases.append(np)
return phrases
tree = ET.parse('ABSA16_Laptops_Train_SB1_v2_with_parse_trees.xml')
@@ -92,16 +110,49 @@ def replace_feature_nps(feature, text, nps):
if len(detected_nps) == 0:
return None
- print(nps)
- print(detected_nps)
unique_nps = list(filter(lambda np: not any(other_np in np for other_np in detected_nps.difference({np})), detected_nps))
- print(unique_nps)
for unique_np in unique_nps:
+ if text == text.replace(unique_np, '$T$'):
+ print('***')
+ print(text)
+ print(nps)
+ print(detected_nps)
+ print(unique_nps)
text = text.replace(unique_np, '$T$')
return text
+# tree_str = '''(ROOT
+# (S
+# (S
+# (NP (PRP I))
+# (VP (MD would)
+# (VP (VB recommend)
+# (NP (PRP it)))))
+# (, ,)
+# (PP (IN for)
+# (NP
+# (NP (NN anybody))
+# (VP (VBG needing)
+# (NP (DT a) (JJ reliable) (JJ simple) (NN laptop)))))
+# (. .)))'''
+# trees = Tree.fromstring(tree_str)
+# for tree in trees:
+# for subtree in tree.subtrees():
+# if subtree.label() == 'NP':
+# np, t = get_np_tree(subtree)
+# print(np)
+# print(t)
+# # print(' '.join(np.leaves()))
+#
+# ns = filtered_np(subtree)
+# print(ns)
+#
+# print('---')
+
+
+
for review in reviews:
sentences = review[0]
assert sentences.tag == 'sentences'
@@ -118,11 +169,11 @@ for review in reviews:
replaced_text = replace_feature_nps(opinion[0], text, nps)
if replaced_text:
prepped_opinions += 1
- print('---')
- print(text)
- print(replaced_text)
- print(opinion)
- print('---')
+ # print('---')
+ # print(text)
+ # print(replaced_text)
+ # print(opinion)
+ # print('---')
else:
pass
# print('---')
--
GitLab