class ClueParser:
def __init__(self):
self.classifier = NaiveBayes()
pass
def findEntities(self, clue):
entities = []
matches = re.findall(location_finder, clue)
for match in matches:
location = match[0] + ", " + match[1]
entities.append(location)
matches = re.findall(entity_finder, clue)
for match in matches:
entities.append(match)
return entities
def findCapEntity(self, clue):
matches = re.findall(capital_finder, clue)
if len(matches) != 0:
return " ".join(matches)
else:
return ""
def parseClues(self, clues):
"""Parse each clue and return a list of parses, one for each clue."""
parses = []
for clue in clues:
klass = self.classifier.classify(self.findFeatures(clue))
entities = self.findEntities(clue)
parse = klass + ":"
if len(entities) != 0:
parse += entities[-1]
else:
parse += self.findCapEntity(clue)
parses.append(parse)
return parses
def findFeatures(self, clue):
words = []
start = 0
i = 0
matches = re.findall(capital_remover, clue)
sent = []
for match in matches:
sent.append(match[0])
clue = "".join(sent)
while i < len(clue):
if clue[i] == "<":
print clue
if (clue[i] == " " or i == len(clue) - 1) and i != start:
word = clue[start:i]
if word[0] == " ":
word = word[1:]
words.append(word)
start = i + 1
i += 1
return words
def train(self, clues, parsed_clues):
"""Trains the model on clues paired with gold standard parses."""
klasses = []
for answer in parsed_clues:
klass = answer[:answer.index(":")]
klasses.append(klass)
features = []
for clue in clues:
features.append(self.findFeatures(clue))
self.classifier.addExamples(features, klasses)
def evaluate(self, parsed_clues, gold_parsed_clues):
"""Shows how the ClueParser model will score on the training/development data."""
correct_relations = 0
correct_parses = 0
for parsed_clue, gold_parsed_clue in it.izip(parsed_clues, gold_parsed_clues):
split_parsed_clue = parsed_clue.split(":")
split_gold_parsed_clue = gold_parsed_clue.split(":")
if split_parsed_clue[0] == split_gold_parsed_clue[0]:
correct_relations += 1
if (split_parsed_clue[1] == split_gold_parsed_clue[1] or
split_parsed_clue[1] == "The " + split_gold_parsed_clue[1] or
split_parsed_clue[1] == "the " + split_gold_parsed_clue[1]):
correct_parses += 1
print "Correct Relations: %d/%d" % (correct_relations, len(gold_parsed_clues))
print "Correct Full Parses: %d/%d" % (correct_parses, len(gold_parsed_clues))
print "Total Score: %d/%d" % (correct_relations + correct_parses, 2 * len(gold_parsed_clues))
class ClueParser:
def __init__(self):
# TODO: if your implementation requires one or more trained classifiers (it probably does), you should declare it/them here.
# Remember to import the class at the top of the file (from NaiveBayes import NaiveBayes)
# e.g. self.classifier = NaiveBayes()
self.classifer = NaiveBayes()
pass
def feature_extractor(self, clue):
"""Given a clue represented as a raw string of text, extract features of the clue and return them as a list or set."""
features = []
words = clue.split()
for word in words:
features.append(word)
special_words = [
"married", "college", "university", "president", "based", "born",
"mayor", "located", "year", "died", "birthplace", "birth"
]
for word in special_words:
if word in clue:
features.append("1")
else:
features.append("0")
return features
def train(self, clues, parsed_clues):
"""Trains the model on clues paired with gold standard parses."""
featuresList = []
relations = []
for clue in clues:
featuresList.append(self.feature_extractor(clue))
for clue in parsed_clues:
relations.append(clue[:clue.find(":")])
self.classifer.addExamples(featuresList, relations)
def getMatch(self, clue, pattern, relation):
matches = re.findall(pattern, clue)
specialWords = [
"is", "was", "married", "is married", "president", "leader",
"head", "founder", "husband", "marriage", "college", "university",
"degree", "wife", "has been", "in charge of", "based",
"headquartered", "headquared", "run", "located", "is in", "led",
"headed"
]
words = []
for m in matches:
if isinstance(m, tuple):
for index, elem in enumerate(m):
if elem not in specialWords and len(elem) > 0:
words.append(elem)
else:
return m
if len(words) > 1 and words[0] != words[1]:
return words[0] + ", " + words[1]
elif len(words) > 0:
return words[0]
return 'no match'
def getEntity(self, relation, clue):
pattern = ""
if relation == "wife_of":
pattern = 'married [\sa-zA-Z]*<PERSON>([\.A-Za-z\s\']*)</PERSON>|wife of [\sa-zA-Z]*<PERSON>([\.A-Za-z\s\']*)</PERSON>|<PERSON>([.A-Za-z\s]*)</PERSON> married|wife is [\sA-Za-z]*<PERSON>([\.\'A-Za-z\s]*)</PERSON>|<PERSON>([\.\'A-Za-z\s]*)</PERSON>\'s [\sA-Za-z]*(wife|marriage)|the wedding between her and [\sA-Za-z]*<PERSON>([\.\'A-Za-z\s]*)</PERSON>|<PERSON>([\.\'A-Za-z\s]*)</PERSON> (is married|married)'
elif relation == "husband_of":
pattern = 'married [\sa-zA-Z<>/]*<PERSON>([\.A-Za-z\s\']*)</PERSON>|husband of [\sa-zA-Z<>/]*<PERSON>([\.A-Za-z\s\']*)</PERSON>|<PERSON>([.A-Za-z\s]*)</PERSON> married|wife is [\sA-Za-z<>/]*<PERSON>([\.\'A-Za-z\s]*)</PERSON>|<PERSON>([\.\'A-Za-z\s]*)</PERSON>\'s [\sA-Za-z<>/]*(husband|marriage)|the wedding between him and [\sA-Za-z<>/]*<PERSON>([\.\'A-Za-z\s]*)</PERSON>|<PERSON>([\.\'A-Za-z\s]*)</PERSON> (is married|married)|<PERSON>([\.A-Za-z\s\']*)</PERSON> met this man'
elif relation == "college_of":
if clue.count("PERSON") > 2:
pattern = '<PERSON>([\.A-Za-z\s\']*)</PERSON>\'s alma mater|<PERSON>([\.A-Za-z\s\']*)</PERSON>[\sa-zA-Z]*(college|university|degree)|alma mater of <PERSON>([\.A-Za-z\s\']*)</PERSON>|to [\sa-zA-Z]*<PERSON>([\.A-Za-z\s\']*)</PERSON>'
else:
pattern = '<PERSON>([\.A-Za-z\s\']*)</PERSON>'
elif relation == "univ_president_of":
if clue.count("ORGANIZATION") > 2:
pattern = "<ORGANIZATION>([&\.A-Za-z\s\']*)</ORGANIZATION> (is|was|has been) [\sa-zA-Z]*(headed|led|founded)|<ORGANIZATION>([&\.A-Za-z\s\']*)</ORGANIZATION>\'s [\sa-zA-Z]*(president|leader|head|founder)|(president|leader|head|founder)[\sA-Za-z<>/\']*of [\sA-Za-z<>/\']*<ORGANIZATION>([&\.A-Za-z\s\']*)</ORGANIZATION>|<ORGANIZATION>([&\.A-Za-z\s\']*)</ORGANIZATION> (president|leader|head|founder)|(headed|led|founded|in charge of) <ORGANIZATION>([&\.A-Za-z\s\']*)</ORGANIZATION>"
else:
pattern = "<ORGANIZATION>([\.A-Za-z\s\'&]*)</ORGANIZATION>"
elif relation == "headquarters_loc":
if clue.count("ORGANIZATION") > 2:
pattern = "<ORGANIZATION>([&\.A-Za-z\s\']*)</ORGANIZATION> [\sa-zA-Z]*(based|headquartered|headquared|run|located|is in)|headquarters of <ORGANIZATION>([&\.A-Za-z\s\']*)</ORGANIZATION>|([&\.A-Za-z\s\']*) is run|([\.A-Za-z\s\'&]*) [\sa-zA-Z]* based|([&\.A-Za-z\s\']*) is in|([&\.A-Za-z\s\']*)'s [\sa-zA-Z]* office|([&\.A-Za-z\s\']*)'s [\sa-zA-Z]* headquarters"
else:
pattern = "<ORGANIZATION>([\.A-Za-z\s\']*)</ORGANIZATION>|([&\.A-Za-z\s\']*) is run|([\.A-Za-z\s\'&]*) [\sa-zA-Z]* based|([&\.A-Za-z\s\']*) is in|([&\.A-Za-z\s\']*)'s [\sa-zA-Z]* office|([&\.A-Za-z\s\']*)'s [\sa-zA-Z]* headquarters"
elif relation == "born_in":
if clue.count("PERSON") > 2:
pattern = "<PERSON>([\.A-Za-z\s\']*)</PERSON>[\sa-zA-Z,<>/]* born [\sa-zA-Z,]*in|birthplace [\sa-zA-Z,<>/]*<PERSON>([\.A-Za-z\s\']*)</PERSON>"
else:
pattern = "<PERSON>([\.A-Za-z\s\']*)</PERSON>"
elif relation == "parent_org_of":
pattern = "(([A-Z][&\.A-Za-z\s\']*)+) (is|was)[\sa-zA-Z<>/1-9,]*offshoot|(([A-Z][&\.A-Za-z\s\']*)+) (is|was) [\sa-zA-Z<>/1-9,]*organization|(([A-Z][&\.A-Za-z\s\']*)+) (is|was) [\sa-zA-Z<>/1-9,]*orginazation|parent [\sa-zA-Z]*<ORGANIZATION>([&\.A-Za-z\s\']*)</ORGANIZATION>|<ORGANIZATION>([&\.A-Za-z\s\']*)</ORGANIZATION> (is|was) [\sa-zA-Z]*offshoot|<ORGANIZATION>([&\.A-Za-z\s\']*)</ORGANIZATION> (is|was) [\sa-zA-Z]*organization|parent \sa-zA-Z]*(([A-Z][&\.A-Za-z\s\']*)+)|ORGANIZATION>([&\.A-Za-z\s\']*)</ORGANIZATION> (is|was) [\sa-zA-Z]*orginazation|parent [\sa-zA-Z<>/]*(([A-Z][&\.A-Za-z\s\']*)+)"
elif relation == "mayor_of":
if clue.count("LOCATION") < 4:
pattern = "<LOCATION>([A-Za-z\s]*)</LOCATION>, ([A-Za-z]*)'s mayor|mayor of <LOCATION>([A-Za-z\s]*)</LOCATION>, ([A-Za-z]*)|<LOCATION>([A-Za-z\s]*)</LOCATION>, ([A-Za-z]*)[\sa-zA-Z<>/1-9,]*is[\sa-zA-Z<>/1-9,]*(led|headed|run) by|mayor of ([A-Za-z]*), <LOCATION>([A-Za-z\s]*)</LOCATION>|([A-Za-z]*), <LOCATION>([A-Za-z\s]*)</LOCATION>[\sa-zA-Z<>/1-9,]*is[\sa-zA-Z<>/1-9,]*(led|headed|run) |mayor of ([A-Za-z]*), ([A-Za-z]*)"
else:
pattern = "<LOCATION>([A-Za-z\s]*)</LOCATION>, <LOCATION>([A-Za-z\s]*)</LOCATION> [\sa-zA-Z<>/1-9,]*is led by|mayor of <LOCATION>([A-Za-z\s]*)</LOCATION>, <LOCATION>([A-Za-z\s]*)</LOCATION>|<LOCATION>([A-Za-z\s]*)</LOCATION>, <LOCATION>([A-Za-z\s]*)</LOCATION>'s mayor|<LOCATION>([A-Za-z\s]*)</LOCATION>, <LOCATION>([A-Za-z\s]*)</LOCATION>[\sa-zA-Z<>/1-9,]* is [\sa-zA-Z<>/1-9,]*(headed|led|run)|<LOCATION>([A-Za-z\s]*)</LOCATION>, <LOCATION>([A-Za-z\s]*)</LOCATION>'s mayor"
elif relation == "univ_in":
pattern = "school [\sa-zA-Z<>/1-9,]*in <LOCATION>([A-Za-z\s]*)</LOCATION>, <LOCATION>([A-Za-z\s]*)</LOCATION>|college [\sa-zA-Z<>/1-9,]*in <LOCATION>([A-Za-z\s]*)</LOCATION>, <LOCATION>([A-Za-z\s]*)</LOCATION>|university [\sa-zA-Z<>/1-9,]*in <LOCATION>([A-Za-z\s]*)</LOCATION>, <LOCATION>([A-Za-z\s]*)</LOCATION>|<LOCATION>([A-Za-z\s]*)</LOCATION>, <LOCATION>([A-Za-z\s]*)</LOCATION> [\sa-zA-Z<>/1-9,]*school|<LOCATION>([A-Za-z\s]*)</LOCATION>, <LOCATION>([A-Za-z\s]*)</LOCATION> [\sa-zA-Z<>/1-9,]*college|<LOCATION>([A-Za-z\s]*)</LOCATION>, <LOCATION>([A-Za-z\s]*)</LOCATION> [\sa-zA-Z<>/1-9,]*university|<LOCATION>([A-Za-z\s]*)</LOCATION>, ([A-Za-z]*) is home|([A-Za-z]*), <LOCATION>([A-Za-z\s]*)</LOCATION> is home|<LOCATION>([A-Za-z\s]*)</LOCATION>, <LOCATION>([A-Za-z\s]*)</LOCATION> in|<LOCATION>([A-Za-z\s]*)</LOCATION>, <LOCATION>([A-Za-z\s]*)</LOCATION> is home|based in <LOCATION>([A-Za-z\s]*)</LOCATION>, <LOCATION>([A-Za-z\s]*)</LOCATION>|located in <LOCATION>([A-Za-z\s]*)</LOCATION>, ([A-Za-z]*)|in <LOCATION>([A-Za-z\s]*)</LOCATION>, ([A-Za-z]*)"
elif relation == "year_of_birth":
pattern = "<PERSON>([\.A-Za-z\s\']*)</PERSON>[\sa-zA-Z<>/1-9,]*(was|is) born|<PERSON>([\.A-Za-z\s\']*)</PERSON>'s birthday|birthday of [\sa-zA-Z1-9,]*<PERSON>([\.A-Za-z\s\']*)</PERSON>"
elif relation == "year_of_death":
pattern = "<PERSON>([\.A-Za-z\s\']*)</PERSON>[\sa-zA-Z<>/1-9,]*died|<PERSON>([\.A-Za-z\s\']*)</PERSON>[\sa-zA-Z<>/1-9,]*passed"
if len(pattern) > 0:
answer = self.getMatch(clue, pattern, relation)
if answer == "no match" and relation != "mayor_of" and relation != "univ_in":
stop_index = clue.find("</")
start_index = clue.find(">")
answer = clue[start_index + 1:stop_index]
if answer == "no match" and clue.count("</") > 1 and (
relation == "mayor_of" or relation == "univ_in"):
stop1 = clue.find("</")
stop2 = clue.find("</", stop1 + 2)
start1 = clue.find(">")
temp = clue.find(">", start1 + 1)
start2 = clue.find(">", temp + 1)
answer = clue[start1 + 1:stop1] + ", " + clue[start2 + 1:stop2]
return answer
return 'Gene Autry'
def parseClues(self, clues):
"""Parse each clue and return a list of parses, one for each clue."""
parses = []
for clue in clues:
feature = self.feature_extractor(clue)
clue_relation = self.classifer.classify(feature)
clue_entity = self.getEntity(clue_relation, clue)
# if clue_entity == "no match":
# print(clue + " " + clue_relation)
parses.append(clue_relation + ':' + clue_entity)
return parses
#### You should not need to change anything after this point. ####
def evaluate(self, parsed_clues, gold_parsed_clues):
"""Shows how the ClueParser model will score on the training/development data."""
correct_relations = 0
correct_parses = 0
for parsed_clue, gold_parsed_clue in it.izip(parsed_clues,
gold_parsed_clues):
split_parsed_clue = parsed_clue.split(":")
split_gold_parsed_clue = gold_parsed_clue.split(":")
if split_parsed_clue[0] == split_gold_parsed_clue[0]:
correct_relations += 1
if (split_parsed_clue[1] == split_gold_parsed_clue[1]
or split_parsed_clue[1]
== "The " + split_gold_parsed_clue[1]
or split_parsed_clue[1]
== "the " + split_gold_parsed_clue[1]):
correct_parses += 1
print "Correct Relations: %d/%d" % (correct_relations,
len(gold_parsed_clues))
print "Correct Full Parses: %d/%d" % (correct_parses,
len(gold_parsed_clues))
print "Total Score: %d/%d" % (correct_relations + correct_parses,
2 * len(gold_parsed_clues))
datapoints = tdatapoints
classes = tclasses
print "reverting to stateless mode"
else:
pickle.dump( datapoints, open( "save.p", "wb" ) )
pickle.dump( classes, open( "save.p1", "wb" ) )
print len(datapoints)
print len(classes)
if('train' in sys.argv[1]):
print 'training done model saved to save.p and save.p1'
sys.exit (0)
nb.addExamples(datapoints,classes)
tests=[]
tests.append(getFeatures('Your graphs were wrong. Please correct it'))
tests.append(getFeatures('thanks john :) I dont get it'))
tests.append(getFeatures("thanks for your answer"))
tests.append(getFeatures('I appreciate your honest response. thanks'))
tests.append(getFeatures('I am not sure how to implement this. Can a TA please clarify? We have been stuck on this problem for quite awhile'))
tests.append(getFeatures("I love them. I can't work on it now. I'm totally confused"))
tests.append(getFeatures("Is the 3 page length a hard limit. It was not clear in the assignment specification"))
tests.append(getFeatures("""
I have no clue about what this class is about.
"""))
tests.append(getFeatures(
