def batchEvaluate(self, parsed_filename: str,
gold_filename: str) -> PARSEVALResult:
parsed_file = open(parsed_filename, 'r')
gold_file = open(gold_filename, 'r')
evaluation_results = list()
while True:
parsed_brackets = parsed_file.readline()
gold_brackets = gold_file.readline()
if not parsed_brackets or not gold_brackets: break
else:
parsed = ParseTree(parsed_brackets)
parsed.makeTree()
gold = ParseTree(gold_brackets)
gold.makeTree()
evaluation_results.append(self.evaluate(parsed, gold))
avgPrecision = 0.0
avgRecall = 0.0
for result in evaluation_results:
avgPrecision += result.precision
avgRecall += result.recall
n = len(evaluation_results)
return PARSEVALResult(avgPrecision / n,
avgRecall / n) # Using macro evaluation
def separate(self, output_train: str, output_test: str, test_instances=-1):
train = open(output_train, 'w')
test = open(output_test, 'w')
count_test = 0
for line in self.treebank:
pt = ParseTree(line)
pt.makeTree()
wordTokens = pt.getWordQueue()
if self.hasOOV(wordTokens):
print(line, file=train, end='')
for token in wordTokens:
self.vocabularies.add(token)
else:
print(line, file=test, end='')
count_test += 1
if test_instances != -1 and test_instances == count_test: break
def __updateScoreText(self, gold_tree: str, parsed_tree: str, idx: int):
GOLD = ParseTree(gold_tree)
GOLD.makeTree()
PARSED = ParseTree(parsed_tree)
PARSED.makeTree()
parseval = PARSEVALEvaluator()
result = parseval.evaluate(PARSED, GOLD)
self.SCORE.configure(text='[' + str(idx) + '] ' + str(result))
self.SCORE.pack(fill='x')
def __process_line(self, brackets):
if brackets[0] == '(':
brackets = brackets[1:len(brackets) - 1]
pt = ParseTree(brackets)
try:
pt.makeTree()
check_result = ParseTreeChecker.__checkTree(pt)
tree_height = pt.getHeight()
if check_result and tree_height > 2:
print(brackets, file=self.accepted)
else:
self.count_rejected += 1
print('[SAD] Found REJECTED:', self.count_rejected)
print(brackets, file=self.rejected)
except ValueError as err:
self.count_rejected += 1
print('[OH NO] Found EXCEPTION:', self.count_rejected)
print(brackets, file=self.rejected)
print('REASON: ', err, file=self.rejected)
def runVisualize(bracket: str, useNLTK: bool):
if useNLTK:
from parsetree.ParseTreeVisualizer import ParseTreeVisualizer
visualizer = ParseTreeVisualizer(bracket)
visualizer.start()
else:
from parsetree.ParseTree import ParseTree
pt = ParseTree(bracket)
pt.makeTree()
pt.printPretty(4, 2)
def getConstituents(self, pt: ParseTree) -> set:
result = set()
# Numbering words (leaf nodes)
def numberingLeaves(pt: ParseTree):
if pt.IS_LEAF:
pt.BEGIN_NUM = numberingLeaves.leaf_num
pt.END_NUM = numberingLeaves.leaf_num
numberingLeaves.leaf_num += 1
numberingLeaves.leaf_num = 0
pt.dfs(numberingLeaves)
# Numbering constituents (non-leaf nodes)
def getBegin(pt: ParseTree) -> int:
pt.BEGIN_NUM = getBegin(
pt.CHILDREN[0]) if pt.BEGIN_NUM == -1 else pt.BEGIN_NUM
return pt.BEGIN_NUM
def getEnd(pt: ParseTree) -> int:
pt.END_NUM = getEnd(
pt.CHILDREN[len(pt.CHILDREN) -
1]) if pt.END_NUM == -1 else pt.END_NUM
return pt.END_NUM
pt.dfs(getBegin)
pt.dfs(getEnd)
# Extract constituents
def extractConstituent(pt: ParseTree):
if not pt.IS_LEAF:
result.add(
PARSEVALConstituentTokens(pt.NODE_TAG, pt.BEGIN_NUM,
pt.END_NUM))
pt.dfs(extractConstituent)
return result
def getBegin(pt: ParseTree) -> int:
pt.BEGIN_NUM = getBegin(
pt.CHILDREN[0]) if pt.BEGIN_NUM == -1 else pt.BEGIN_NUM
return pt.BEGIN_NUM
def numberingLeaves(pt: ParseTree):
if pt.IS_LEAF:
pt.BEGIN_NUM = numberingLeaves.leaf_num
pt.END_NUM = numberingLeaves.leaf_num
numberingLeaves.leaf_num += 1
def getEnd(pt: ParseTree) -> int:
pt.END_NUM = getEnd(
pt.CHILDREN[len(pt.CHILDREN) -
1]) if pt.END_NUM == -1 else pt.END_NUM
return pt.END_NUM