Uning tuzilishini aniq va quyi oqim algoritmlari tomonidan oson boshqarish mumkin

Ob'ektlarni aniqlang (grafikdagi tugunlar)
Bizning ta'rifimizga ko'ra, alohida so'zlar grafikdagi tugunlardir. Shunday qilib, hujjatdagi har bir noyob so'z uchun tugun bo'lishi kerak .
def get_entities(document):
# in our case, entities are all unique words
unique_words = []
for sent in document:
for word in sent:
if word not in unique_words:
unique_words.append(word)
return unique_words
Aloqalarni aniqlang (grafikdagi tugunlar orasidagi qirralar)
Bizning ta'rifimizga ko'ra, agar tegishli so'zlar kirish hujjatining jumlasida bigramma hosil qilsa, grafikdagi ikkita tugun o'rtasida chekka mavjud . Yaxshiyamki, dastlabki ishlov berish bosqichida biz hujjatni mos ravishda ajratamiz, bu esa ushbu hisob-kitobni osonlashtiradi.
def get_relations(document):
# in our case, relations are bigrams in sentences
bigrams = []
for sent in document:
for i in range(len(sent)-1):
# for every word and the next in the sentence
pair = [sent[i], sent[i+1]]
# only add unique bigrams
if pair not in bigrams:
bigrams.append(pair)
return bigrams


Grafikni qurish
Endi bizda tugunlar va qirralar bor, biz haqiqiy grafikni osongina qurishimiz mumkin. Ushbu misolda men grafik ob'ektni yaratish uchun NetworkX dan foydalanaman . Endi biz oddiy funktsiya yordamida haqiqiy grafikni chizishimiz mumkin.
def plot_graph(G, title=None):
# set figure size
plt.figure(figsize=(10,10))
# define position of nodes in figure
pos = nx.nx_agraph.graphviz_layout(G)
# draw nodes and edges
nx.draw(G, pos=pos, with_labels=True)
# get edge labels (if any)
edge_labels = nx.get_edge_attributes(G, 'weight')
# draw edge labels (if any)
nx.draw_networkx_edge_labels(G, pos, edge_labels=edge_labels)
# plot the title (if any)
plt.title(title)
plt.show()
return

Build_graph() asosiy funksiyasini ishga tushirgandan so‘ng namunaviy hujjatning yo‘naltirilmagan, vaznsiz grafik syujeti. Muallif tomonidan tasvir.