def find_hearst_concepts(self, triples):
s_concepts = []
m_concepts = []
for (t1, rel, t2) in triples:
term1 = Term(preprocessor.pos_tag(t1, True))
term2 = Term(preprocessor.pos_tag(t2, True))
synsets1 = wn.synsets(term1.get_head()[0], self.pos_tag(term1.get_head()[1]))
synsets2 = wn.synsets(term2.get_head()[0], self.pos_tag(term2.get_head()[1]))
if not synsets1:
raise Exception("'{}' not found in WordNet".format(term1.get_head()[0]))
if not synsets2:
raise Exception("'{}' not found in WordNet".format(term2.get_head()[0]))
(best1, best2) = self.comp(synsets1, synsets2)
con1 = self.get_concept(
concept.Concept(synset=best1, term=term1.get_head()[0])
)
con2 = self.get_concept(
concept.Concept(synset=best2, term=term2.get_head()[0])
)
conChild1 = None
conChild2 = None
if len(term1.get_terms()) > 1:
conChild1 = self.get_concept(
concept.Concept(name=term1.get_terms(), term=term1.get_head()[0])
)
con1.add_hyponym(conChild1)
conChild1.add_hypernym(con1)
#m_concepts.append(conChild1)
if len(term2.get_terms()) > 1:
conChild2 = self.get_concept(
concept.Concept(name=term2.get_terms(), term=term2.get_head()[0])
)
con2.add_hyponym(conChild2)
conChild2.add_hypernym(con2)
#m_concepts.append(conChild2)
if conChild1:
if conChild2:
conChild1.add_relation(conChild2, rel)
else:
conChild1.add_relation(con2, rel)
m_concepts.append(conChild1)
else:
if conChild2:
con1.add_relation(conChild2, rel)
m_concepts.append(conChild2)
else:
con1.add_relation(con2, rel)
s_concepts.append(con1)
s_concepts.append(con2)
self.single_concepts = self.single_concepts.union(set(s_concepts))
self.multi_concepts = self.multi_concepts.union(set(m_concepts))
def load(book):
Definitions = []
Propositions = []
suf = bookToRoman(book)
base_url = "http://aleph0.clarku.edu/~djoyce/java/elements/"
url = base_url + "book" + suf + "/" + "book" + suf + ".html"
page = requests.get(url)
soup = BeautifulSoup(page.text)
for conceptType in soup.find_all('dl'): #[0].find_all('dd'):
prev = conceptType.previous_element.previous_element
if prev == 'Definitions' or prev == "Common Notions" or prev == "Postulates":
i = 1
for dd in conceptType.find_all('dd'):
Definitions.append(
concept.Concept(
prev[0:3], book, i,
dd.get_text().replace('\r', '').replace('\n', ""), "",
[], []))
i += 1
if prev == 'Propositions':
count = 0
for a in conceptType.find_all('b'):
if "Proposition" in str(a):
count += 1
for i in range(1, count + 1):
url = base_url + "book" + suf + "/prop" + suf + str(
i) + ".html"
page = requests.get(url)
soup = BeautifulSoup(page.text)
print url
statement = soup.find_all('div',
class_='statement')[0].get_text()
text = ""
fromLink = []
for p in soup.find('div', class_='theorem').find_all('p'):
text += p.get_text()
for j in soup.find('div',
class_='theorem').find_all('div',
class_='just'):
for l in justificationToLink(j.get_text()):
fromLink.append(l)
Propositions.append(
concept.Concept(
"Pro", book, i,
statement.replace('\r', '').replace('\n', ''),
text.replace('\r', '').replace('\n', ''), [],
fromLink))
return [Definitions, Propositions]
def form(self, terms):
#actual formation
sets = []
multiwords = []
#look for multiword terms, for concepts of them and then go on with the disambiguation only
#with the head-term
for term in terms:
if len(term.get_terms()) > 1:
con = self.get_concept(
concept.Concept(name=term.get_terms(), term=term.get_head()[0])
)
multiwords.append(con)
synsets = self.lookUp(term.get_head()[0], term.get_head()[1])
if synsets:
sets.append(synsets)
if sets:
easies = self.get_easies(sets)
rest = [x for x in sets if not x in easies]
concepts = self.compare_easies(easies)
concepts = self.compare_concepts(concepts, rest)
for (mult, con) in itertools.product(multiwords, concepts):
if str(con.get_term()) == str(mult.get_term()):
mult.add_hypernym(con)
con.add_hyponym(mult)
self.single_concepts = set(concepts)
self.multi_concepts = set(multiwords)
def compare_concepts(self, concepts, rest):
#Compares already found concepts to possible synsts for each term
#in order to find most probable one
conNoTerm = [x[0] for x in concepts]
restNoTerm = [y[0] for y in rest]
for synsets in restNoTerm:
similarities = []
for possib in synsets:
similarity = 0
for con in conNoTerm:
similarity = similarity + possib.path_similarity(con)
similarities.append(similarity)
concepts.append((synsets[similarities.index(max(similarities))],
rest[restNoTerm.index(synsets)][1]))
result = []
for conc in concepts:
c = self.get_concept(
concept.Concept(synset=conc[0], term=conc[1], name=conc[1])
)
result.append(c)
return result
def concept_from_formula(fmla):
vs = sorted(list(ilu.used_variables_ast(fmla)), key=str)
name = (','.join(str(v) + ':' + str(v.sort) for v in vs) + '.' + str(fmla))
return co.Concept(name, vs, fmla)